Fix arctic_hydrology climatology Blaskey-adjusted JSON vs. original GCM CSV discrepancies

[cstephen]

This PR fixes the discrepancy in values between outputs of these two URLs (for example):

http://localhost:5000/arctic_hydrology/modeled_climatology/81009008
http://localhost:5000/arctic_hydrology/modeled_climatology/81009008?format=csv

The problem was that the gcm_diff_applied_to_blaskey adjustment was being done after the CSV export code, not before, for both the modeled_climatology and wt_modeled_climatology endpoints.

I've also added a couple chunks of code to move the "historical" model above the projected models in the CSVs to make them more intuitive and consistent with our other CSVs.

To test, I pointed the hydroviz webapp at this json_csv_discrepancy Data API branch and disabled LOWESS smoothing + enabled hovermode to cross reference values from the hydrographs & water temperature hydrographs with their corresponding CSVs to make sure that the same values appear on the same dates. I also cross-referenced a handful of flow & temperature stat table values against their corresponding CSVs. Everything is as expected!

[charparr]

I've tested the JSON and CSV routes for a few different IDs, both CONUS and AK, and can verify they are aligned in terms or structure, numerical content, and associated metadata.

[charparr]

Note the test failures here are just referencing missing env vars for kuti, so can disregard.