error collection/recovery and reporting improvements

[zkat]

Fixes: #93

the v2 parser can do this now:

[epage]

Didn't expect this to be more literally resume_after but only for cuts. I take it FromRecoverableError is working out? Any other input on the rest of the error recovery design?

FromRecoverableError and Recoverable just felt like they were getting messy and I wasn't too pleased with the results.

[zkat]

honestly, once I figured out the right "surgical" cut I wanted (literally just resume_after but for cuts), things started chugging along nicely and got a lot more intuitive.

The problem with the current resume_after is that, a lot of times when I want to resume, it's when I specifically get an unambiguous syntax error ("I was definitely parsing a number, but I ran into something bad, and there's no way it could be anything else. abort abort"), but if it could be something else, I want my alts to keep working!

Besides that, FromRecoverable and Recoverable work... mostly fine, except it was a little confusing to figure out where error-related stuff should go (for example, you have to apply context before the resume_after, otherwise the collected error never gets the context applied, which is obvious in hindsight but was a bit tricky to figure out. Docs would probably be good enough here). I'm sure they can be made simpler, but functionality-wise, they do what's needed, I think.

[epage]

The problem with the current resume_after is that, a lot of times when I want to resume, it's when I specifically get an unambiguous syntax error ("I was definitely parsing a number, but I ran into something bad, and there's no way it could be anything else. abort abort"), but if it could be something else, I want my alts to keep working!

Its easy for me to forget how much other people rely on alt. In toml_edit, I think my use if dispatch! makes it so if I resume_after in specific places, I don't have to worry about backtracking.

The intention behind including backtracking is to deal with making sure all errors that can be recovered are as I likely don't have good enough coverage in cut_err to ensure resume_after_cut will always work. I guess you could always cut_err and resume_after_cut. The docs would need to explicitly call out that workflow so people are less likely to overlook it.

Besides that, FromRecoverable and Recoverable work... mostly fine, except it was a little confusing to figure out where error-related stuff should go (for example, you have to apply context before the resume_after, otherwise the collected error never gets the context applied, which is obvious in hindsight but was a bit tricky to figure out. Docs would probably be good enough here). I'm sure they can be made simpler, but functionality-wise, they do what's needed, I think.

Thanks for that feedback!