ListObjects: don't recurse beyond prefix

[lunixbochs]

This fixes the biggest performance issue with ListObjects for me - it would recurse into all subdirectories, even when using a delimiter.

There are still a few issues:

• Large directories take forever to list.
• Parent directories are still walked for no reason - perhaps the posix backend should scope the os.DirFS() to the last directory in the prefix rather than starting at the root, and stitch the paths back together in p.fileToObj. I looked at doing that but wasn't sure if I'd hit a prefix = "" edge case.

Also fixes #903

[versity-github]

This won't automatically run in continuous integration without approval. A member of the Versity organization must allow it.

[benmcclelland]

set a const in this file to "/" and use that instead of using "/" everywhere

[lunixbochs]

I have mixed feelings about this, it reduces readability, fs.FS is never going to use something other than a slash, and several other files in this repo also hardcode "/"

[lunixbochs]

The new logic should work with any delimiter:

if delimiter != "" &&
    strings.HasPrefix(path+"/", prefix) &&
    strings.Contains(strings.TrimPrefix(path+"/", prefix), delimiter) {
    skipflag = fs.SkipDir

If the delimiter is non-empty, and our path is beyond the prefix, and the delimiter is still in the directory path after trimming the prefix from it, we shouldn't recurse.

I fixed the pagination issue by not doing an early return, instead opting to set a skipflag variable which is used instead of return nil for the rest of the walk closure.