rm: implement iterator interface

[tertsdiepraam]

Part of nushell/nushell#10447 (comment)

This path makes an iterator interface for rm. The iterator is the Remover type, which yields an UResult<()> for every file. nushell can choose to how to print the errors.

To achieve that, I had to make every function return errors instead of printing them. This in turn lead to some big refactors for remove_file and remove_dir. I wish there was a better way to write an iterator like this, but I think it turned out alright. Maybe we can clean it up more in the future.

Oh and walkdir was only getting in the way. So I removed that too.

[tertsdiepraam]

Looks like I got some tests to fix 😄

[github-actions]

GNU testsuite comparison:

Congrats! The gnu test tests/rm/ir-1 is no longer failing!
GNU test failed: tests/pwd/pwd-long. tests/pwd/pwd-long is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/rm/deep-2. tests/rm/deep-2 is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/rm/inaccessible. tests/rm/inaccessible is passing on 'main'. Maybe you have to rebase?
GNU test failed: tests/rm/unread2. tests/rm/unread2 is passing on 'main'. Maybe you have to rebase?

[sylvestre]

@tertsdiepraam sorry but it needs some rebasing

[fdncred]

any updates here?

[tertsdiepraam]

@fdncred Oof, it's been a while. I might have some ideas to bring this forward. An iterator may not be the right approach here. The code was getting too complicated.

But, I think we have missed the easy solution here: calling remove multiple times with a single file. I think that's much easier for us to build and for nushell to use. If I recall correctly, that should be enough for nushells's usecase.

[tertsdiepraam]

I'm gonna go with a 2 step plan here:

1. Take some of the refactors I did and extract them into a separate PR.
2. And then re-evaluate the error handling. Either with a channel or an iterator or a callback kind of thing.