Bias correction fix

[benowen-bom]

Addresses #1968.

This PR updates the way masks are inherited from the forecast and truth cubes to the forecast_error. Rather than happening implicitly (as is currently the case), the propagation of masks is now performed explicitly.

Unit and acceptance tests were also updated to considered masked data (for both forecast and truth) to ensure the masks are passed on as expected.

The updated acceptance test data is included in metoppv/improver_test_data#36.

Testing:

• Ran tests and they passed OK
• Added new tests for the new feature(s)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (3eee0a9) 98.40% compared to head (1b5c73c) 98.40%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1969   +/-   ##
=======================================
  Coverage   98.40%   98.40%           
=======================================
  Files         124      124           
  Lines       12028    12030    +2     
=======================================
+ Hits        11836    11838    +2     
  Misses        192      192           

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

[dmentipl]

Thanks @benowen-bom this addresses #1968. The addition of tests checking for masked arrays vs non-masked arrays is good. I have checked that the acceptance tests pass locally. I understand that @bomRob has run a suite on a workstation to check that the results are as expected.

[benowen-bom]

Thanks @benowen-bom this addresses #1968. The addition of tests checking for masked arrays vs non-masked arrays is good. I have checked that the acceptance tests pass locally. I understand that @bomRob has run a suite on a workstation to check that the results are as expected.

Thanks @dmentipl for the review!