Add fieldvector unit tests, update rcompare

[charleskawczynski]

This PR adds a fieldvector unit test, which was extracted from #2070, where the unit tests passed but a test in ClimaLand broke.

It's still not clear to me whether this is exercising broadcasting over different fieldvectors types with the same propertynames, or something different. Either way, we should probably test for that, plus, we could update rcompare to handle fieldvectors with different propertyname orders.

[charleskawczynski]

I think updating rcompare to handle different propertyname orders is okay, since users can easily test typeof(x) == typeof(y), but I figure we can make these a keyword option so that users can pick which they want. By default, though, the == will be strict, I think that's probably the safest option.

[charleskawczynski]

This fails when combined with #2070, so this should help us figure out what is wrong with #2070.

[charleskawczynski]

Ok, I think I've identified what is being exercised: nested fieldvectors with different propertyname ordering. I'll update the PR