Make sure signed/unsigned in our protocol handling is correct

[roji]

We had some confusion around the parameter number limit actually being unsigned in the protocol (#2795). This likely means we have other bugs in this area; specifically, message lengths are currently treated as int, although they probably need to be uint.

[roji]

Am moving this to 5.0 as this is probably a non-trivial change and we haven't had any actual complaints about it.

[roji]

@Brar am moving this to the backlog. While it's good for us to proactively do this at some point, we can also decide to close this and react to any error reports as they come, your call.

[NinoFloris]

During both attempts at fixing GHSA-x9vc-6hfv-hg8c I have carefully reviewed all our message writing code, changed what wasn't ideal and found the rest to be correct. Final optional improvements can be found in #5734