Review test coverage #9433

Description

While working on issue #8995 I wondered whether there is sufficient test coverage for the binary (FITS, BINARY and BINARY2) serializations. The non-binary serialization (TABLEDATA) tests seem to have good coverage. I would like to look at whether those tests should be replicated for the binary cases, or whether the existing tests of the converters are sufficient to show that the various data types and shapes (array values) are being handled consistently across the different serializations.

In addition, PR #9505 made it clear that we could be more complete in covering the subtle aspects of VOTable format and semantics. This is important because some aspects of the VOTable spec are subject to interpretation, and more test cases could demonstrate the understanding or assumptions that were made in the implementation.

Additional context

This was inspired by #8995 and may be related to #8737.