First take at removing HeaderError from YaccParser internals

[ratmice]

This is a first take at removing the Location/Header error from the internals of Parser,
I haven't gone over the diff extremely carefully.

My original intent was to make Header generic on either Span or Location, and then use Header<Span> inside, and Header<Location> in CTParserBuilder and nimbleparse.

What I noticed though, that it was possible to do it without that change, by returning a Result<..., HeaderError> from what is essentially YaccParser::new rather than from parse where we currently return YaccGrammarErrors.

We absolutely could still make Header generic, the parsed header results all return Location::Span.
I thought it was worth noting this before I went ahead and did it, since the generic Span type permeates through both the errors, and the Value type.

[ltratt]

I think this one is ready now?

[ratmice]

I think it is ready, but I'd like to read through the patch more closely which I won't have time to do until tomorrow most likely.

[ratmice]

Likely needs #[doc(hidden)], and I should just go through the docs again, to ensure there isn't anything else I added that also needs it.

[ratmice]

I believe all the doc changes should be fixed across various commits,
this one in 7895675

[ratmice]

I did manage to get my self-review done before heading out which was mostly just documentation stuff,
but also hiding HeaderError from the public facing API, that allowed basically hiding the entire module from top-level.

I suppose I'm happy with this now, the public facing API seems good at least.

The callers/header building and what not seems like it could probably use some refinement,
but that all seems well within the scope of stuff that can be done as we come across things that can be improved (primarily error messages).

[ltratt]

Please squash.

[ratmice]

Is that alright? (It leaves the unwrap() in the first patch to avoid squashing those two fairly large patches together)

Edit: I suppose it would be nice if we moved the test movement/#[cfg(test)] to YaccGrammar to the first patch, but that is also changing the type of it's return value currently, so it would need some patch surgery.

Edit 2x: I guess it is at least worth a try.

[ratmice]

I went ahead and switched moving the tests and added the FromStr implementation to the first patch
(modified to temporarily use HeaderError and unwrap)

That removes the #[cfg(test)] function entirely from the series, but that means adding a second call to unwrap in the first patch which then gets fixed in the second patch.

[ltratt]

Is that alright? (It leaves the unwrap() in the first patch to avoid squashing those two fairly large patches together)

Yep!