Document that `ser_float_to_uint32` is not the inverse of `ser_uint32_to_float` #21964

Document that ser_float_to_uint32 is not the inverse of ser_uint32_to_float.

Document that ser_double_to_uint64 is not the inverse of ser_uint64_to_double.

Document that the only supported round-trip is: floating-point → integral → floating-point.

Context: #21929 (comment)

<!--e57a25ab6845829454e8d69fc972939a-->

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

<!--174a7506f384e20aa4161008e828411d-->

Conflicts

Reviewers, this pull request conflicts with the following ones:

• #21981 (Remove unused float serialization by MarcoFalke)
• #21966 (Software float encoding by sipa)

If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.

See also #21966 for an alternative that makes float/double serialization exactly specified and independent of platform.

Closing in favour of #21966 which removes the need for this documentation :)