refactor uu_ls so that crate users can call the ls without having to print everything to stdout

[fdncred]

I was wanting to use uu_ls in nushell and found out that I couldn't because some of the internals weren't exposed in a way such that crate callers could really use it. So, I was wondering if something like this PR would be something worth considering?

It refactors the code in such a way so that crate callers could call uu_ls and get structured data back instead of everything printing to stdout. It does this without a breaking change by implementing a type of visitor pattern through the LsOutput trait.

I created a TextOutput struct that impelments the LsOutput trait so now the list() function now looks like this:

   pub fn list(locs: Vec<&Path>, config: &Config) -> UResult<()> {
       let mut output = TextOutput::new(config);
       list_with_output(locs, config, &mut output)
   }

All of this allows crate callers like nushell to now do something like this:

use uu_ls::{Config, list_with_output, CollectorOutput};
use std::path::Path;

let mut output = CollectorOutput::new();
list_with_output(vec![Path::new(".")], &config, &mut output)?;

for entry in output.entries() {
    // Convert to Nushell Value with full type information
    let name = entry.display_name.to_string_lossy();
    let size = entry.size().unwrap_or(0);
    let is_dir = entry.is_dir();
    // ...
}

One last note, I know I didn't discuss this with anyone so it won't hurt my feelings if this PR isn't accepted. I'm open to whatever. I was just trying to figure out a way to reuse more of the excellent coreutils code. Thanks for such a great repo!!!

[github-actions]

GNU testsuite comparison:

Note: The gnu test tests/pr/pr-tests was skipped on 'main' but is now failing.

[sylvestre]

sorry, it needs to be rebased