fix(ls): respect sorting when grouping directories; disambiguate size sorting properly

[Alonely0]

Fixes #11997.

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/timeout/timeout-group. tests/timeout/timeout-group is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/date/date-locale-hour (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/date/resolution (fails in this run but passes in the 'main' branch)

[cakebaker]

Can you please add a test to ensure we don't regress in the future?

[Alonely0]

Are the other instances in 7bfa92f OK?

It is because we do it for the first time on the buffer and only once; there stability doesn't matter. Let me add a test real quick.

[Alonely0]

That should be it. I had to also fix the size checking in order to get the new tests to pass (and that wasn't my fault LOL).

[cakebaker]

Thanks!

[yurikhan]

So sorting by time will also put equivalently-timestamped entries in random order?

[Alonely0]

We could also disambiguate those by name, yeah. Kinda surprising we previously left it in the order std gave it to us, although I have to say that this is essentially impossible outside fabricated scenarios or no{a, c, m}time filesystems. I'll look into it as well.

[yurikhan]

I was raised on filesystems that only tracked one timestamp field and on software whose all files, at normal conditions, had a modification date/time of 1992-03-10 03:10:00. I know fabricated scenarios to be a fact of life, not a remote possibility.