fs: utf8 fast paths don't accept all valid utf8 values

[CanadaHonk]

UTF8 and UTF-8 are also valid but not accepted by any existing utf8 fast paths (#48658, #49884)

[Uzlopak]

Seems there needs to be a call to normalizeEncoding

[CanadaHonk]

I don't think that would be needed, there's only 4 cases where that returns utf8 and it's with the values: utf8, utf-8, UTF8, UTF-8. It might be worth making a small util to check for those (not sure if the function call would add overhead or not).

[Uzlopak]

yeah, but you could also have mixed cases, like 'Utf-8' right?

[CanadaHonk]

Oh right as a final check. I don't think fast paths should have to worry about that for now at least as I doubt it's common and would add measurable overhead?

[Uzlopak]

I once wrote for testing all permutations of uppercase/lowercase into a Map and it improved the performance of normalizeEncoding. This sounds like a use case for normalizeEncoding.