Use quantile delta mapping (QDM) to bias adjust forecast, reanalysis

[brews]

This PR adds prototype quantile-delta mapping (QDM) bias adjustment of ERA5 and season forecast ensembles to GMFD before mortality projection.

To run with this change:

1. Download forecast ensemble with scripts/00-download_forecast.py. This now downloads multiple years' worth of forecast ensembles initialized at a particular month. The data is downloaded to a local directory. Currently, this is 1981 - 2026 for May. I let this run overnight. It downloads under 70 GiB of data.

2. Run scripts/01-parse_forecast.py. This parses the downloaded data and writes its output to POREALLAS_PARSED_FORECAST_URI.

3. Run scripts/01-parse_era5.py and scripts/01-parse_gmfd.py. I run this on a daskhub cluster hosted on Google Cloud. It uses ERA5 data from Google's ARCO store (as before) and the CIL's internal GMFD data store. These write output to POREALLAS_PARSED_ERA5_URI and POREALLAS_PARSED_GMFD_URI, respectively.

4. Run scripts/02-bias_adjust_era5.py and scripts/02-bias_adjust_forecast.py. These use the data parsed above as input. Output is written to POREALLAS_ERA5_URI and POREALLAS_TAS_FORECAST_URI, respectively. Bias adjustment for ERA5 takes about 3 minutes on my laptop. Bias adjustment of the forecast ensemble takes roughly 2.5 hours.

5. Now you can project mortality as you would have before by running scripts/05-project_effects.py. This writes full, "hot only" and "cold only" projections of the forecast and baseline period.

A few additional things to note:

• Parsed files have their calendar converted to "noleap" and so output data files now also use a "noleap" calendar. This was needed because the QDM implementation needs input data to use the same calendar. Pandas and geopandas might have trouble using this calendar. You can convert it back to a standard calendar for analysis with Xarray's convert_calendar.
• All the output Zarr Stores in these scripts use consolidated metadata.

The README and example.env have been updated to reflect these changes.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 69.79%. Comparing base (3ef4ab8) to head (e018cb9).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #17   +/-   ##
=======================================
  Coverage   69.79%   69.79%           
=======================================
  Files           3        3           
  Lines          96       96           
=======================================
  Hits           67       67           
  Misses         29       29           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[brews]

After talking with the group, I'm holding off on merging this because it has breaking changes to #16. Otherwise, this should be ready to go, AFAIK.

I need to switch to work on another project, and I'll be out for a week+ on Friday. I'm moving the assignment to @kemccusker so she has the power.