Fix nml object use

[hcorson-dosch-usgs]

Fixes #22

When I started to dig into #22, I realized that it was possible to make this fix without writing each of the nml objects to file, so I instead implemented an objects-based fix here.

This PR:

• Adds list iteration to the p1_nml_objects list of nml objects
• Maps over that p1_nml_objects in the target recipe for p2_glm_uncalibrated_runs, but does so as part of a cross() with p1_gcm_names and p1_gcm_dates, which means that each nml object is repeated 18 times - once for each gcm-time-period combination. That matches the construction of p1_model_config, which means that the nml object passed to each branch of p2_glm_uncalibrated_runs is the correct nml object for that model run.

I tested this out for a subset of 5 lakes. If I first run the model for 4 of the lakes, and then add another site_id for a 5th lake, only the models for that 5th lake are built. Similarly if I run the model for 4 of the lakes and then remove one of those site_ids, none of the models are rebuilt. That's the behavior we want, and corrects the issue noted in #22.

[hcorson-dosch-usgs]

Also wanted to note that I tested the mapping by doing the following:

1. Subsetting p1_lake_cell_tile_xwalk to 5 lakes:

   tar_target(p1_lake_cell_tile_xwalk_df, 
            readr::read_csv(p1_lake_cell_tile_xwalk_csv, col_types=cols()) %>%
              filter(site_id %in% p1_nml_site_ids) %>%
              arrange(site_id) %>%
              filter(site_id %in% c('nhdhr_105567868','nhdhr_105569520', 'nhdhr_114336097',
                                    'nhdhr_120019185','nhdhr_114544667'))),

2. Passing p1_gcm_names and p1_gcm_dates to the run_glm3_model() command:

   tar_target(
   p2_glm_uncalibrated_runs,
   run_glm3_model(
     sim_dir = '2_run/tmp/simulations',
     nml_obj = p1_nml_objects,
     gcm_name = p1_gcm_names,
     gcm_date = p1_gcm_dates,
     model_config = p1_model_config,
     burn_in = 300,
     burn_out = 190,
     export_fl_template = '2_run/tmp/GLM_%s_%s_%s.feather'),
   packages = c('retry','glmtools', 'GLM3r'),
   pattern = map(p1_model_config, cross(p1_nml_objects, p1_gcm_names, p1_gcm_dates))),

3. Modifying run_glm3_model to return a tibble to check that the pertinent parameters match:

   run_glm3_model <- function(sim_dir, nml_obj, gcm_name, gcm_date, model_config, burn_in, burn_out, export_fl_template) {
   # pull parameters from model_config
   site_id <- model_config$site_id
   time_period <- model_config$time_period
   gcm <- model_config$gcm
   raw_meteo_fl <- model_config$meteo_fl
   
   mapping_check <- tibble(
   config_site_id = site_id,
   nml_site_id = nml_obj$morphometry$lake_name,
   config_gcm = gcm,
   mapped_gcm = gcm_name,
   config_time_period = time_period,
   mapped_time_period = gcm_date
   )
   return(mapping_check)

   In the resulting tibble, you can see that the mapping set up by cross(p1_nml_objects, p1_gcm_names, p1_gcm_dates) matches the order of p1_model_config: