Scaling GCM munge steps to full footprint

[lindsayplatt]

Fixes #279 (but not #280). Ended up with tiles of 20x20 cells. With these updates, the full build took just over 15 hours (the download piece alone was 13, which was less than what it was for just MN locally).

Looking for @hcorson-dosch to review:

• _targets.R
• 7_drivers_munge.yml
• 7_drivers_munge/out/7_GCM_driver_files.ind
• build/status/N19kcml2ZXJzX211bmdlL291dC83X0dDTV9kcml2ZXJfZmlsZXMuaW5k.yml

Looking for @jesse-ross to review all the others (which he mostly has already):

• README.md
• .gitignore
• docker/*
• gcm_drivers.slrum
• gcm_download.R (though, I would really like this to just become a line in the slurm script, but I don't know how)
• launch-rstudio-container.slurm

-- Late additions below this line --

Forgot to include this quick little code snippet for how I determined the time required to run the sub-pipeline.

calc_minutes <- function(metadata) {
  round(sum(metadata$seconds, na.rm = TRUE)/60, 2)
} 

tar_meta(starts_with('gcm_data_raw_feather')) %>% calc_minutes() # Took 13.9 hours to download
tar_meta(gcm_nc) %>% calc_minutes() # Took 17 minutes to convert to NetCDFs.
tar_meta() %>% calc_minutes() # The full pipeline took 15.2 hours

[jesse-ross]

@lindsayplatt I set up the .sif file on tallgrass, and created the symlink with ln -s lake-temperature-model-prep_v0.1.sif lake-temperature-model-prep.sif, so once you pull the changes down on tallgrass, you should be ready to submit to SLURM! We may need to iterate on the container, so please let me know if it runs into problems.

[lindsayplatt]

@jesse-ross thank you! Can't wait to try it out