Fix LocaleIterator for platforms where c_char is u8 (better fix?)

[geniot]

I had the same problem as the previous pull-requester when compiling on aarch64:

error[E0308]: mismatched types
  --> /root/.cargo/git/checkouts/rust-sdl2-4bcf6005ccaf1a39/ecd03de/src/sdl2/locale.rs:78:33
   |
78 |     let region = try_get_string(sdl_locale.country);
   |                  -------------- ^^^^^^^^^^^^^^^^^^ expected `*const i8`, found `*const u8`
   |                  |
   |                  arguments to this function are incorrect
   |
   = note: expected raw pointer `*const i8`
              found raw pointer `*const u8`
note: function defined here
  --> /root/.cargo/git/checkouts/rust-sdl2-4bcf6005ccaf1a39/ecd03de/src/sdl2/locale.rs:86:11
   |
86 | unsafe fn try_get_string(ptr: *const i8) -> Option<String> {
   |           ^^^^^^^^^^^^^^ --------------

error[E0308]: mismatched types
  --> /root/.cargo/git/checkouts/rust-sdl2-4bcf6005ccaf1a39/ecd03de/src/sdl2/locale.rs:90:39
   |
90 |         Some(std::ffi::CStr::from_ptr(ptr).to_string_lossy().into_owned())
   |              ------------------------ ^^^ expected `*const u8`, found `*const i8`
   |              |
   |              arguments to this function are incorrect
   |
   = note: expected raw pointer `*const u8`
              found raw pointer `*const i8`
note: associated function defined here
  --> /rustc/05f9846f893b09a1be1fc8560e33fc3c815cfecb/library/core/src/ffi/c_str.rs:264:25

Anyway, I believe my fix is a better one... Choose one :)

[Cobrand]

I do think this fix is better than the other one, but please add an entry to the changelog because this will definitely break those already using this with aarch64.

[geniot]

I do think this fix is better than the other one, but please add an entry to the changelog because this will definitely break those already using this with aarch64.

Change log updated.

What do you mean "this will definitely break those already using this with aarch64" ?
On the contrary, this change fixes SDL2 usage on aarch64.

[Cobrand]

Ah right my bad, disregard my comment about the breaking change then, I see now it can't compile on aarch64 without this fix, I thought it compiled but the output type was not user-friendly.