mruby-string-ext: validate UTF-8 sequences with mrb_utf8len - #7105

mrb_utf8len() decides what counts as a UTF-8 character in core, under the
RFC 3629 rules it was given in #7093. mruby-string-ext decides it twice more,
in utf8code() and in str_scrub_char_len(), and both were taught those same
rules by hand in that same PR. Ask core instead.

Reading a refusal

mrb_utf8len() answers 1 for every sequence it turns down. A real one-byte
character is a lead byte below 0x80, so a length of 1 on a byte at or above
0x80 is the refusal. That is the whole bridge; both helpers are a few lines on
top of it.

static mrb_int
str_scrub_char_len(const unsigned char *p, const unsigned char *e)
{
  mrb_int len = mrb_utf8len((const char*)p, (const char*)e);
  if (len == 1 && p[0] >= 0x80) return -1;
  return len;
}

utf8code() keeps only the part core has no answer for, the code point itself,
and reports the byte length it decoded through a new lenp argument.
String#codepoints was stepping over characters by the raw
mrb_utf8len_table, which is the length a lead byte claims rather than the
length that was accepted; it now steps by what came back. The table is still
read by String#chars, so the declaration stays.

22 lines of C replace 50.

Equivalence

Nothing here is meant to change behavior, so the old and new helpers were run
against each other over both leading bytes, nine representative tail bytes and
every available length from 1 to 4: 21,233,664 cases, with no difference in
the code point, in what is refused, or in the byte length.

Tests

String#ord and String#codepoints had nothing holding them to the RFC 3629
rules; the tests for those go through String#scrub, which reads its lengths
from the other helper. The first commit pins them, and it passes both before
and after the two that follow.

Verified

• rake test on a full-core build (MRB_UTF8_STRING through mruby-encoding):
  2243 tests, all green, at each of the three commits
• rake test on the default gembox (no MRB_UTF8_STRING): 2055 tests, all
  green
• prek run --all-files passes, except that markdownlint could not install
  locally (npm engine mismatch); no Markdown is touched here

Summary by CodeRabbit

• Bug Fixes

  • Improved UTF-8 decoding for string operations.
  • String#ord and String#codepoints now correctly reject malformed UTF-8 sequences with ArgumentError.
  • UTF-8 scrubbing now consistently validates character sequences.

• Tests

  • Added coverage for malformed UTF-8, including invalid bytes following valid four-byte characters.