Improve record encoding code

[bugadani]

This PR is one step in the direction of a write buffer. The intention is to introduce a builder that can encode records in write memory, then encrypt them if necessary, in-place. This PR moves most of the record encoding and touch-up into a single place, not counting encrypted application data building. The PR also cleans up weird and unnecessary return values.

[bugadani]

I'm wondering what the original design decisions were and if I'm breaking something (even possible compatibility with earlier TLS) very subtly with my refactors. I understand that I'm moving stuff around a lot, but I promise it's not completely pointless, most likely, maybe :)

[bugadani]

I'm not noticing anything here that would break that, could you point me to to what piece of code you're thinking about?

Code organization in general seems to have special cases (like the post-processing of the client hello) that make me think of "this must have been done for some reason I don't understand". Nothing in particular, though.

with a lot of conflicts

Sorry about that 🙈

[lulf]

From what I can see, this particular special case (TLS-PSK) is still supported with the change. Wrt. the code organization, I think you're the first that actually are trying to refactor and clean stuff after the initial release, so as features and improvements have took place there was no refactoring to put things in a better structure. So, thanks a lot for that effort, I can see it improves by every PR!

[bugadani]

From what I can see, this particular special case (TLS-PSK) is still supported with the change

I do my best trying to not delete features :)