Basic parser error recovery

[bamarsha]

Add support for error recovery to the scanner, as well as some primitive parsers for ease of use. Include some basic error recovery for items and statements.

[idavis]

Does it make sense to add recovery for statements missing ; as part of this PR or as a follow up?

[bamarsha]

Can you be more specific? This PR does recover from statements missing ;. It recovers by eating all following statements until it finds a ;, which is not the smartest way to recover, but does avoid breaking the whole file. Do you mean you think the recovery should be smarter in this PR?

[idavis]

Yes, smarter. Resume parsing as if ; had been there. Change the range of the error to where the ; should have been. The current impl has the error span on the next statement start.

[bamarsha]

The diagnostic error spans are unchanged by this PR, it's the same as on main.

I think there are enough changes for infrastructure and refactoring in this PR that I don't want to spend too much time on the recovery behavior itself. I just did basic recovery for items and statements as a proof of concept for the recovery API. There's a long tail of recovery behavior improvements that we can do in follow-up PRs.

[idavis]

Sounds good. I'll finish reviewing the rest of the PR.