Switch to debiased GCMs

[lindsayplatt]

David has those available now on THREDDS. Fixes the data but also adds longwave radiation, which will need to be a new variable added to our workflow.

• [x] Switch to downloading the new data (now daily and you can download 1980-2099 all at once)
• [x] Check to see if these two issues (timezone woes & relative humidity conversions) are still needed
• [x] Verify the grid projections are correct (see discussion here) (completed, read comments in this issue below)
• [x] Check that the dates (and missing Dec 31sts) are OK (see details in this comment) Fixed by #299
• [x] (moved to its own issue) #296
• [x] Munged feathers in target glm_ready_gcm_data_feather sometimes have NAs and sometimes have NaNs. How come? This causes an issue with building our NetCDF file, so we added a fix that converts NAs to NaNs to get put into the NetCDF file.

[lindsayplatt]

Now functioning as expected! I forgot to use the correct column for one of my steps, which created some funkiness. So, in Winslow, all the leap years ended on Dec 30 but Feb 29 was included. In our data, I have adjusted it so that no Feb 29ths are included, but all the years end on Dec 31.

[lindsayplatt]

The final mystery in this issue was solved! We wanted to explore why some variables have NAs and some had NaNs. All the missing data from GDP had NaNs. During the processing step, any column that we did some math on (such as Rain and Snow), ends up converting the NaNs to an NA. That's all there was to it! We are going to keep our fix within the NetCDF creation that converts those NAs back into NaNs for storage purposes.