Remove implicit conversions from String, Char16String and CharString to data pointers.

[Ivorforce]

This is mostly a sanity change to avoid accidental conversions leading to unexpected behavior.

I have not tested how many callers are affected (e.g. String s = char_string).
All that are will experience a noticeable improvement of speed because strlen does not have to be called.

Notes on implicitness

I have made conversions from String-likes to StrRange implicit to aid the transition.
I thought long and hard about whether I want to 'contribute' another implicit conversion to the codebase, seeing as though they are already leading to quite a lot of problems.

However, I think conversion to StrRange is one of the rare cases where it actually is warranted:

• The conversion does not lose metadata about the contents (pointer and length).
• No data is allocated, and the conversion is trivial.

The only gotcha for this conversion is really that it may lead to ambiguous calls, and it may lead to unintentional loss of ownership. However, both are true about the const char * conversions, so it's an upgrade all around.

Behavior changes

There are a few behavior changes that may lead to problems (though they are unlikely to). If problems are experienced somewhere, it's likely the caller should have been more explicit from the start, and should be addressed to do so.

Especially, callers will change behavior when encountering a \0: Instead of cutting the string short, they will add the character to the middle of the string. In terms of UTF-32, the new behavior is correct. Still, if unexpected behavior is observed, it should be evaluated on a case-by-case basis - most likely, the previous behavior was incorrect other behavior should be adjusted to fit.

[Ivorforce]

Sorry about the repeated pushes!! I should really learn to test this kinda change on my own repo to get other platforms to pass before I make a PR 🙃

[bruvzg]

I assume this is superseded by #101304, which include all changes from this PR.

[Ivorforce]

I have reset two calls I had changed from latin1 parsing to utf8 parsing. It was incorrect to change them; they should be assessed on a case by case basis. If I haven't missed anything all calls I changed to call get_data() explicitly will now behave the exact same as before.

I assume this is superseded by #101304, which include all changes from this PR.

Yes! Although I would be happier if this PR was merged / approved first because it has some implications that are not discussed in the #101304 separately.

[akien-mga]

I assume this is superseded by #101304, which include all changes from this PR.

The commit here is a pre-requisite in #101304, but it would be best to approve and merge this first, and then rebase #101304, to keep concerns separate.