fix: wrong handling of values and timesteps in temporal models

[frodehk]

Why is this pull request needed?

Problem when extending time series for compressors across temporal models in EnergyModelBaseResult.extend(). This is basically related to how None-values are treated. Two bugs:

• Loss of data: When first temporal model contains None and next temporal model contains real values, the merge/extend function evaluates to None - ignoring data from second temporal model.
• Mismatch between timesteps and values: When first temporal model contains real values, and next temporal model contains None, the merge/extend function evaluates to the values of the first temporal model. The timesteps in the second model are ignored, hence a mismatch between number of timesteps and number of values are introduced for the actual time series.

What does this pull request change?

• CompressorTrainResultSingleStep: Ensure that all time series have values or NaN for all valid timesteps, even if input is None
• CompressorModelSampled: Ensure that all time series have values or NaN for all valid timesteps, even if input is None
• Update snapshots
• Add test

Issues related to this change:

https://equinor-ecalc.atlassian.net/browse/ECALC-376?atlOrigin=eyJpIjoiNzY4MzVlMWU2ZjEzNDMwNzhiYzNhMDZkMmExYzljOTQiLCJwIjoiaiJ9

[TeeeJay]

it looks good, but all snapshots are expanding null to NaN..I guess we have no cases where real values are actually overwritten? Good with test. Need UAT to have a look to QA this :) Good job though :)

[frodehk]

it looks good, but all snapshots are expanding null to NaN..I guess we have no cases where real values are actually overwritten? Good with test. Need UAT to have a look to QA this :) Good job though :)

Yes, I think None is resolved to null in json (?). And, None is now filled with nans to ensure correct number of values vs timesteps.