Ensure parse_utf8 has length of string passed in when available

[kiroxas]

As seen in #99826, parse utf_8 is faster when it doesn't have to find the size of the string first.
So if calling code has it, make sure to pass it along.

[Ivorforce]

It might be that this is changing behavior; current code is terminating on NULL even if the actual string length was larger. Not that i'm again it; just noting. The tests will probably see if it still works.

[kiroxas]

It might be that this is changing behavior; current code is terminating on NULL even if the actual string length was larger. Not that i'm again it; just noting. The tests will probably see if it still works.

It should not, as length is just a hint to be able to allocate a destination buffer large enough to hold the result ( which is 1 to 1 if the file is just ASCII, or could be 4 to 1 if only 4 bytes codepoints). It is only used in the conversion algorithm as a stop condition, and if we iterated past the length, it was a bug, as the case I changed allocated length bytes just before the call. If it stopped on null, we will stop on null too, even if it's smaller than length. The loop condition is the same in current implementation and the PR.

[Repiteo]

Thanks!