boulder/moss: Support shell completions generation

[ReillyBrogan]

This wires up clap_complete so that the completions subcommand of boulder and moss will now automatically generate shell completions when provided a shell as the single argument.

The second commit replaces the alias implementation in moss with one that works better with how Clap works with shell completions. Now commands like moss ur correctly show further completions as though the user had used moss repo update

Closes #220

[ikeycode]

OK I'm liking in principle but I don't actually know how I feel about matching on None as a magic value for completions?

[ReillyBrogan]

OK I'm liking in principle but I don't actually know how I feel about matching on None as a magic value for completions?

The reason I'm doing that is in order to support moss li as an alias for moss list installed we have to attach the command for installed to the root command object as li (but hidden so it doesn't appear in --help). Ordinarily when moss list installed is called we unpack the clap return as a subcommand list which has a further child installed. When we call it as moss li instead we just get the equivalent of the final installed object. So that's why the pattern matcher matches on None currently, if moss li is called then there's no further subcommand to extract out and route with so we need to use the additional context that's passed to the install handler.

An alternative would be to split up the "figure out what subcommand was called" logic from the "route to appropriate handler" logic which would probably make it look a bit cleaner.

[ermo]

An alternative would be to split up the "figure out what subcommand was called" logic from the "route to appropriate handler" logic which would probably make it look a bit cleaner.

Per Ikey's comments, this is probably a good idea, assuming you're up for it?