Bug fix: Ensure proper handling of errors during partial data definitions

[RHammond2]

This PR fixes a series of small bugs in the validation logic of PlantData:

• Ensures that no asset geometry parsing occurs when PlantData.asset is None
  • Adds PlantData.calculate_asset_geometries() for users that add asset data post-validation
• Ensures reanalysis non-definition does not error out the validation steps by adding in a default "product" key for the metadata definition
• Adds a step that ensures all of the reanalysis data products all have matching metadata definitions, or an error occurs early in the initial data loading/validation stage
• Ensures that when None is in PlantData.analysis_type in addition to other values, that the error raising step is not skipped

Closes #233

Thank you @jordanperr for identifying that there were a series of related issues when only some data is provided to PlantData!

[jordanperr]

I validated this PR by reading the code, merging it with entralliance/OpenOA/from_entr_connectorentralliance:feature/from_entr_constructor and running the electrical losses and AEP notebooks, and also running the plantdata_validator regression tests locally. It looks good to me.

Tests are failing due to known issue https://github.com/NREL/OpenOA/issues/228