Input functions should take Option

[willmurnane]

This PR contrasts with PR 641. The primary motivation here is that the various input functions take optional arguments, and produce optional results. This is implemented by changing the interface in pgx/src/inoutfuncs.rs, then the macro in pgx-macros/src/lib.rs, then teaching the SQL generator that this is acceptable in pgx/src/datum/mod.rs. Finally, I changed all the implementations of those functions that I could find, so that cargo test --workspace passes with these changes.

[workingjubilee]

I think this needs at least one test to verify that the NULL -> NULL and NULL -> error message cases both work correctly. I am hoping the former is already effectively tested for (haven't checked yet) so it would be the latter, mostly. The latter just to make sure that the Rust &str we return from these functions is correctly used by Postgres during erroring.

[willmurnane]

I added a pair of tests, and fixed up the interface in a minor way (using a constant rather than a function, so that implementors can override it).

[workingjubilee]

Well, an implementation can override a default impl of a function too, no? I think it's mostly about ergonomics and if e.g. we would want to offer dynamic messages (I can't really think of why or how we would, however, since it's just a simple in/out).

[workingjubilee]

This fixes some erroneous behavior and is fully compatible with #615. If any final retouch edits disrupt it I will resolve them myself, though I am still going to land #615 first.

[workingjubilee]

Looks like this passes tests!