[BUG]: Order of bands / tweakback / astrometric align leads to problems with any bands before the one used for registration

[akleroy]

I'm reducing a full MIRI set with 770, 1000, 1130, and 2100 finding that the AstrometricAlignStep runs in fixed band order. This means that if 770 is indicated to inherit its solution from 1000 then that solution is not in place when 770w runs and as a result 770w does not get a shift and (more important?) no tweakback files are generated.

I could be doing something wrong but the issue seems repeatable, I am not sure why it didn't affect Cycle 1's nominal final run - maybe something changed?

An annoying but easy fix that avoids having to do anything smart would be to split out a subsequent "AstrometricInheritStep" to go after the catalog-based alignment step.

[akleroy]

FWIW I confirmed that swapping F770W into the pole position results in all bands getting tweakbacks and works.

[thomaswilliamsastro]

Should be fixed in #110 @akleroy, if you could test that'd be great