Merge WRF-PartMC PartMC specific development to PartMC

[jcurtis2]

This pull request merges in many of the changes I've made for using PartMC with WRF.

Changes included:

• Sources contributing to a particle are tracked as an array aero_components which is replacing n_orig_part to reduce memory requirements. WRF-PartMC for example had 70+ sources which made n_orig_part very large yet sparsely populated as most particles have few sources.
• Weight class selection for each aerosol source to allow intermediate options between not weighting sources separately and representing all sources separately. With 70+ sources in WRF-PartMC, coagulation becomes very slow so sources can be grouped together to reduce costs.
• Option to not sample particles that are a certain number of standard deviations from the mean using a sampling timescale.
• Changed to requiring netCDF4 to allow for int64 for aero_id.
• New preprocessor flags for WRF-PartMC specific variables and processes (PMC_USE_WRF)

Things to consider/address before merging:

• [x] Change particle_error in MOSAIC.F90 to be WRF-PartMC only which avoids MOSAIC killing a run for a "bad" particle.
• [x] Determine what happens to aero_component for per_particle coagulation. aero_component in this case currently gives less information than the old method.
• [x] Should we bring back source weighting as a default so users do not have to specify weight classes in input files.
• [x] Add an integer to aero_particle that tracks the true number of components
• [x] Mixing state size selected - replace current method with making a sampled aero_state
• [x] Fix optical properties to be an array which will be size 1 in the box model and size 5 in WRF-PartMC
• [x] Add various WRF-PartMC subroutines into PartMC
  • [x] partmc_trans_aero.F90: subroutine aero_state_duplicate_particles... added as aero_state_dup_all_particles
  • [x] partmc_trans_aero.F90: real(kind=dp) function aero_state_group_class_num_conc
  • [x] partmc_driver.F90: function aero_state_num_concs_by_source
  • [x] partmc_driver.F90: real(kind=dp) function aero_state_total_num_conc_wet
  • [x] Check aero_state functions that have grid but might not need them. move to PartMC if they do not need grid.

[jcurtis2]

Added some test coverage to check to make sure results are similar regardless of chosen weighting scheme. The following weighting options are considered: NUMMASS_SOURCE, FLAT, NUMASS, FLAT_SOURCE, FLAT_SPECIFIED, NUMMASS_SPECIFIED.