Migrate away from gumdrop for argument parsing

[str4d]

The last gumdrop release was almost two years ago, and its developer has marked their status on GitHub as "On indefinite hiatus from all programming work". In particular, this means that gumdrop is stuck on syn 1 (and it is one of only two dependencies stuck there). So I'm looking to migrate away from gumdrop to another crate.

My initial feeling is that I'll migrate to clap, for a few reasons:

• clap is actively maintained.
• Back when I chose gumdrop over clap, the latter was much heavier with many dependencies. This has apparently improved in recent releases.
• I use clap to generate the shell completion files, so I already have a clap representation of the CLIs.

However, I do want to take this opportunity to examine what other alternatives exist, and what would work best for rage:

• Argument handling must support the same semantics that rage relies on for its flags.
• I don't mind the text UI changing a bit (wrt how flags are displayed in help text etc), but it should have a similar UX (if not better).
• Integration with my localization stack (fluent and i18n-embed) would be really nice. Currently I adjust the help text to have localized explanatory text and examples, but the help text for the actual flags is un-localizable.
• I'm not too concerned about binary size or compile times, but they ideally won't regress too far.

[str4d]

Potentially useful comparision data: https://github.com/rosetta-rs/argparse-rosetta-rs

[str4d]

Reading clap-rs/clap#380 it sounds like although there is no built-in support yet, clap does have sufficient APIs to enable localization, and other people have tested doing so with Fluent. Combined with the fact that I already have a clap representation of the CLIs, I think I'll put effort into that direction and see how far I get.

[str4d]

#442 migrates to clap, which made it pretty easy (once I figured out the kinks) to localize the usage and help text. I didn't end up using any of the existing builder-based clap representation, as migrating from gumdrop to clap's derive-based representation ended up being simpler. I still need to figure out how to migrate the completions etc. away from my current approach to de-duplicate, but if that works then this is almost certainly what I'll merge.