Specimin's modularity model does not match NullAway's with respect to fields

[kelloggm]

NullAway's rules for fields require that all non-null fields be initialized before a constructor completes. The modularity model that NullAway is using here differs from the one that Specimin currently implements (based on the Checker Framework's typechecking rules) in the following ways:

• all nonnull fields in a class can impact whether a warning is issued in a constructor, and so should be preserved. This is because the absence of an assignment to a field can trigger a warning.
• the bodies of private or final methods that are directly called at the top level by a constructor can impact whether NullAway issues a warning in the constructor, because NullAway permits methods with those restrictions to also assign nonnull fields (to prevent the error above)

I think this means we need to introduce a separate modularity model for NullAway that reflects these differences.

This is the reason that Specimin currently does not preserve na-323, even with the change in #331.

[kelloggm]

Partially fixed (first bullet point) by #345.