Adds complete stochastic uncertainty capabilities to CostMethods with supporting changes elsewhere

[rjhanes]

How the CostMethods worked before

• Cost models/methods could be run with no uncertainty, array-based uncertainty using specific values from the scenario YAML file, or stochastic uncertainty using draws from a probability distribution. Parameters can have either no uncertainty or uncertainty within a scenario, so we can explore results variability due to one parameter, a subset of all our parameters, or every parameter. Uncertain parameters within a scenario must all have either array or stochastic uncertainty.
• When the stochastic uncertainty was used, triangular distributions were defined for uncertain parameters and random draws from these distributions were taken every time the cost method was called.
• This is undesirable behavior because we're examining the impact of particular cost method parameters (slope/rate of change and y-intercept/initial value, mostly), and taking draws at every time step doesn't give us coherent cost trends and therefore doesn't give us coherent insight.

How the CostMethods will work following this PR

• When using stochastic uncertainty, a single draw from a parameter's probability distribution is taken at the beginning of each model run (at year 2000) and used throughout that run. Subsequent runs will use additional draws from the distribution, one draw per run.
• The array-based uncertainty capability has not changed. The previously developed "[x]_[param]_oaat" scenario YAMLs will still work with this code. I updated the format for all 20 OAAT YAMLs on OneDrive and tested this PR on the updated 10_fgic_oaat with no errors and expected results. The YAML format updates involved removing the probability distribution parameters, which weren't being used in the OAAT analysis anyway.
• The no uncertainty capabilities are unchanged save for format updates to the scenario YAML - see below.

Data and config changes to support this PR

• No changes to the input datasets or files.
• No changes required to the case study YAML.
• The cost_uncertainty dictionary in the scenario YAML file changed format to allow for stochastic capabilities.
• A scenario YAML for executing 10 stochastic runs on the tiny data set is below, and also stored as a separate file in OneDrive under uncertainty/stochastic-yamls
• See the stochastic-uncertainty branch in the data repo for an updated scenario YAML file that executes a baseline (MCiC-MC) run using this new format. I'll merge those changes in to all our stable data branches after this PR is merged in.

Stochastic scenario YAML for use with tiny data

flags:
  clear_results         : True # If results files already exist, move them to a sub-directory to avoid overwriting
  compute_locations     : True  # if compute_locations is enabled (True), compute locations from raw input files (e.g., LMOP, US Wind Turbine Database)
  run_routes            : True  # if run_routes is enabled (True), compute routing distances between all input locations
  use_computed_routes   : True  # if use_computed_routes is enabled, read in a pre-assembled routes file instead of generating a new one
  initialize_costgraph  : True  # create cost graph fresh or use an imported version
  location_filtering    : False  # If true, dataset will be filtered to the states below
  distance_filtering    : False # if true, filter computed routes based on max distances in route_pairs file
  pickle_costgraph      : True  # save the newly initialized costgraph as a pickle file
  generate_step_costs   : True # set to False if supply chain costs for a facility type vary regionally
  use_fixed_lifetime    : True # set to False to use Weibull distribution for lifetimes
  use_lcia_shortcut     : True # set to False to re-generate the lca_db file

scenario:
    name: Tiny Stochastic Test
    capacity_projection: StScen20A_Fake_annual_state.csv
    states_included:
    seed: 13
    electricity_mix_level : state
    runs: 10

circular_pathways:
    sc_begin:
    - manufacturing
    sc_end: 
    - landfilling
    #sc_in_circ:
    sc_out_circ:
    - cement co-processing
    - next use
    learning:
        coarse grinding:
            component : blade
            uncertainty: 
            initial cumul: 1.0
            cumul: 
            learn rate: -0.05
            steps:
            - coarse grinding
            - coarse grinding onsite
        fine grinding:
            component : blade
            uncertainty:
            initial cumul: 1.0
            cumul: 
            learn rate: -0.05
            steps:
            - fine grinding
    cost uncertainty:
        landfilling:
            uncertainty: stochastic
            m:
                value: 1.5921
            b:
                c: 0.5
                loc: 23.12
                scale: 11.56
                value: 
        rotor teardown:
            uncertainty:
            m: 1467.08
            b: 285.0
        segmenting: 
            uncertainty:
            b: 27.56
        coarse grinding onsite:
            uncertainty: stochastic
            initial cost:
                c: 0.5
                loc: 53
                scale: 106
                value: 
        coarse grinding: 
            uncertainty: stochastic
            initial cost:
                c: 0.5
                loc: 53
                scale: 106
                value: 
        fine grinding: 
            uncertainty: stochastic
            initial cost:
                c: 0.5
                loc: 71.5
                scale: 143
                value: 
            revenue:
                c: 0.5
                loc: 136.5
                scale: 273
                value: 
        coprocessing:
            uncertainty:
            b: 10.37
        segment transpo:
            uncertainty: stochastic
            cost 1:  # Before 2001; 2002-2003
                c: 0.5
                loc: 2.175
                scale: 4.35
                value: 
            cost 2:  # 2001-2002; 2003-2019
                c: 0.5
                loc: 4.35
                scale: 8.70
                value: 
            cost 3:  # 2019-2031
                c: 0.5
                loc: 6.525
                scale: 13.05
                value: 
            cost 4:  # 2031-2044
                c: 0.5
                loc: 8.7
                scale: 17.40
                value: 
            cost 5:  # 2044-2050
                c: 0.5
                loc: 10.875
                scale: 21.75
                value: 
        shred transpo:
            uncertainty: stochastic
            m: 
                c: 0.5
                loc: 0.00056105
                scale: 0.0011221
                value: 
            b: 
                c: 0.5
                loc: 0.0262
                scale: 0.0524
                value: 
        manufacturing:
            uncertainty:
            b: 11440.0
    path_split:
        fine grinding:
            fraction: 
                value: 0.3
            facility_1: landfilling
            facility_2: next use  
        pass:
            next use
    permanent_lifespan_facility: 
    - landfilling
    - cement co-processing
    - next use
    vkmt : 
    component mass : 
    year : 

technology_components:
    circular_components:
    - blade
    component_list:
        nacelle : 1
        blade : 3
        tower : 1
        foundation : 1
    component_materials:
        nacelle : 
        - steel
        blade : 
        - glass fiber
        - epoxy
        tower : 
        - steel
        foundation : 
        - concrete
    component_fixed_lifetimes: # Years
        nacelle : 30
        blade : 20
        foundation : 50
        tower : 50
    component_weibull_params: #L, K
        nacelle : 
        blade : 
            L : 240
            K : 2.2
        foundation : 
        tower :
    substitution_rates:
        sand: 0.15
        coal: 0.30

[rjhanes]

@jwalzberg and @TJTapajyoti A heads up that I'll be going over this PR in our hackathon tomorrow. I've prepared a scenario YAML file to start our stochastic runs on HPC (in OneDrive under uncertainty/stochastic-yamls), and this PR covers changes to the Cost Methods that will enable those runs.

I've identified eight parameters to include in the stochastic runs, based on our OAAT results:

• landfill fee initial value
• fine grinding revenue
• fine grinding initial cost
• coarse grinding initial cost
• coarse grinding onsite initial cost
• segment transport
• shred[ded material] transport rate of change
• shred transport initial value

To ensure that we get results worth interpreting, I recommend we use the +/- 50% range to define the triangular probability distributions. How many runs we execute is up for debate - we should probably discuss how the single-run time has changed with adding two states and the pyLCA 2.0 upgrade and go from there. Since we're not trying to do statistical analyses or actual sensitivity analyses I think we can pretty much pick a nice round number.