Update usage parser (potentially with nom or chomp)

[kbknapp]

With nom recently reaching 1.0 I'm wondering if using would be any benefit to this project. Particularly in the test-ability, performance, and maintainability.

Because I'm very unfamiliar with it, but slowly reading through it, this may be slow going.

Perhaps initially evaluating on the from_usage parser.

To be clear, in order to be merged into mainline it needs to not impact performance at all (which I don't believe will be an issue), and must improve at least one of the items above.

[Vinatorul]

From the one side - we can rewrite parser no nom and make it more clear.
From the other - nom is new dependency.

As for me - I'm pretty much for rewriting :)

[kbknapp]

Yeah I understand the new dep thing, but if we can make the code more readable, concise, testable, and performant then I'm all ok using a new dep. If we can do all the above by just re-writing the parser I'm good with that too. But I would imagine handwriting a new parser might be re-implenting things already solved innom

Also chomp looks promising too but the fact that it's less established and pre-1.0 means I'd lean towards nom all things being equal.

To be clear, I'm not really advocating using nom in master yet, more just testing to see if it's something we should start using.

[kbknapp]

In all honesty using an iterator in the usage parser is silly, but I guess it made sense to me when I did it. We'd probably get way better performance using a real parser, which is what I'd really like to explore. Whether that's using something like nom or home rolled? Let's try them both and see what looks/performs best! 😄

[kbknapp]

Closed with #386