Fix compiler warnings from utf8.c on 32-bit build

[t-a-k]

This set of changes will silence compiler warnings from Perl_utf8_to_uv_msgs_helper_ function at utf8.c on 32-bit (e.g. x86) build:

utf8.c:1819:35: warning: comparison of integer expressions of different signedness: ‘ssize_t’ {aka ‘int’} and ‘unsigned int’ [-Wsign-compare]
 1819 |             && UNLIKELY(expectlen > OFFUNISKIP(uv)))
      |                                   ^
utf8.c:1895:45: warning: comparison of integer expressions of different signedness: ‘unsigned int’ and ‘ssize_t’ {aka ‘int’} [-Wsign-compare]
 1895 |                 for (unsigned i = curlen; i < expectlen; i++) {
      |                                             ^
utf8.c:2098:34: warning: format ‘%d’ expects argument of type ‘int’, but argument 2 has type ‘U32’ {aka ‘long unsigned int’} [-Wformat=]
 2098 |                 Perl_croak(aTHX_ "panic: Unexpected case value in "
      |                                  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 2099 |                                  " utf8n_to_uvchr_msgs() %d", this_problem);
      |                                                               ~~~~~~~~~~~~
      |                                                               |
      |                                                               U32 {aka long unsigned int}

---

• These changes should not change the behavior of perl, so I think that it does not require a perldelta entry.

[mauke]

Why can't this be negative?

[tonycoz]

It can be negative in non-DEBUGGING builds, where s0 <= send is treated as a reported error. For DEBUGGING builds perl asserts.

[t-a-k]

Thank you for comments. I've forgotten a possibility of s0 > send.
I pushed a new commit to move this subtraction to the point after checking s0 < send to avoid underflow.

[jkeenan]

Per suggestion from @mauke on #p5p: We can compare the logs from two different CI runs, one without the patch in this p.r., one with it.

https://github.com/Perl/perl5/actions/runs/12667183737/job/35300550360#step:7:64

vs.

https://github.com/Perl/perl5/actions/runs/12674636619/job/35324408185?pr=22899#step:7:64

This suggests that the patch has the intended effect. However, I can't evaluate the code per se, and @mauke's concern above should be answered.