Draft: feat(strings): Remove depricated string functions

[dark0dave]

@FredrikLindgren this removes as much as possible any old functions which are called by weidu. Should reduce the noise in the build stage too.

[FredrikLindgren]

This is unfortunately one of those types of changes I'm going to be difficult about. Since, say, uppercase is not semantically equivalent to uppercase_ascii it is not necessarily backward-compatible to replace the former with the latter. For all I know, there are mods out there that rely on the case transformation being done in Latin1 and would break otherwise.

There definitely are cases where it is safe to switch: the ones where the input is known, say, up-casing MODDER arguments, and in those cases I'm not arguing. But if the case-transform is on unknown input it's a different matter, for which I feel there is not necessarily a good answer. At the very least, I'd do these in a different commit, so the mod-facing changes could be reverted in case of reports of mod breakage.

I'm not suggesting you should do this, but in the alternate universe in which I actually accomplish anything, I would have combined this review with a critical eye towards whether or not the case-transform was necessary or could be done some other way, especially wrt. linux compatibility (for instance, making a biff up-cases the contents your key-file which is a no-no). Edit: not so much actually doing these types of changes, but charting where work is necessary.

[dark0dave]

@FredrikLindgren ah, I didn't consider that! You are right! My apologies. Would you be open to making a function instead? Just so we can get closer to a new version of ocaml?

[dark0dave]

Some thing like:

let latin1_to_upper (s: string) : string =
  String.map (fun c ->
    if c >= '\x61' && c <= '\x7A' then
      char_of_int (int_of_char c - 32)
    else
      c
  ) s

[FredrikLindgren]

So it seems like mine was perhaps a bit of a void concern. I don't know the inner workings of, say, String.uppercase and String.uppercase_ascii, but it seems like they behave identically on Latin-1 input, that is to say, they just uppercase a-z and leavé any code pöints in the non-ASCII range alone. There ought to be some difference between the two, but I can't find any documentation that retains String.uppercase, and empirical evidence is thus far to the contrary. Still, I would feel better if there was one commit that did this change for known inputs and another that handle the cases where the input is unknown. I can do it, as it is much to ask, but then it's on a more of a glacial time scale.