Add values_dtype backend option to load values at full precision

[juntyr]

Supersedes #395

#395 (comment)

Hi @juntyr and @dmey, thanks for this change! However, I think this is something we would prefer to be an option since it will double the memory requirement (and we do have GRIB fields with over 1 billion values).

Potential API:

ds = xr.open_dataset('data.grib', engine='cfgrib', backend_kwargs={'values_dtype': 'float64'})

Do you think you could make this modification to the PR, or would you rather leave it as a request issue?

[juntyr]

@iainrussell What are your thoughts on the updated PR?

[juntyr]

Ping @iainrussell

[juntyr]

Cc @SF-N

[iainrussell]

Hi @juntyr, sorry for the long delay, but I've got back to this now - your PR looks good, I just added a couple of tests and plan to make a release tomorrow.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 84.00000% with 4 lines in your changes missing coverage. Please review.

Project coverage is 41.85%. Comparing base (11acc76) to head (21ec8e8).

Files with missing lines	Patch %	Lines
cfgrib/dataset.py	0.00%	3 Missing ⚠️
cfgrib/messages.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #407      +/-   ##
==========================================
+ Coverage   41.34%   41.85%   +0.50%     
==========================================
  Files          26       26              
  Lines        2201     2222      +21     
  Branches      194      194              
==========================================
+ Hits          910      930      +20     
- Misses       1291     1292       +1     

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

[juntyr]

Thank you very much for your help @iainrussell!

[iainrussell]

cfgrib 0.9.15.0 has been released with this new feature - thank you @juntyr!