Variance array resampling differs from jwst

[braingram]

spacetelescope/jwst#8344 described issues and proposed updates for resampling variance arrays in jwst.

spacetelescope/jwst#8437 updated jwst with one of the proposed solutions.

The variance array resampling in romancal doesn't appear to have similar updates and is instead using the older version of the variance array resampling.

[schlafly]

Thanks Brett. FWIW, I think in Roman we're happy to follow Webb here, e.g., as part of future refactoring to use more common code in stcal. In discussions with Stefano in the past I don't think RTB likes our current variance computations or the new Webb ones and prefers yet another option based on the weight images, but we haven't coded that up and it's only been requested insofar as I've mentioned that the current variance computation seems incorrect to me.