string.c: search bytes in String#byterindex and in String#rindex on a binary string - #7099

A backward search reads a string one way while the forward search reads it
another. str_index() scans bytes, but str_rindex() walked character
boundaries with char_adjust() / char_backtrack(), and both String#rindex
and String#byterindex went through it. Two pairs of methods disagree as a
result, and neither disagreement depends on malformed input.

byterindex against byteindex

byterindex reports byte positions, so it has to be able to name a position
inside a multi-byte sequence. Walking characters, it never visited one:

s = "aあb"
s.byteindex("\x81")   # => 2
s.byterindex("\x81")  # => nil, CRuby answers 2 for a byte-indexed subject

str_byterindex() now scans one byte at a time, the mirror of str_index(),
and backs String#byterindex. str_char_rindex() keeps the boundary walk and
backs String#rindex, where a match starting inside a character must be
passed over so the search continues to the character before it. The existing
assert_nil broken.rindex("\x81") in test/t/string.rb pins that half down
and still holds. A non-UTF-8 build reaches only the byte scan, so the
char_adjust() and char_backtrack() identity macros defined there are gone.

rindex against index on a binary string

rindex dispatches to the byte search when the single-byte flag is set, but a
binary string that has not had its flag computed went on to the UTF-8 path, so
the walk moved over character boundaries and a negative position was counted
in characters. index counts bytes for the same string:

s = "aあb".b
s.index("\x81".b)      # => 2
s.rindex("\x81".b)     # => nil, CRuby answers 2
s.rindex("\x81".b, -2) # => nil, CRuby answers 2
s.length               # marks the string single-byte
s.rindex("\x81".b)     # => 2

Adding RSTR_BINARY_P to that dispatch reads the string the way
mrb_str_char_to_byte() and mrb_str_byte_to_char() already read it. Every
value above now matches CRuby 4.0.6, before and after the flag is set.

Testing

• rake test with build_config/host-debug.rb (full-core, MRB_UTF8_STRING):
  mrbtest 2235 OK / 0 KO, bintest 116 OK.
• rake test with the default config (byte strings): mrbtest 2033 OK / 0 KO,
  bintest 105 OK.
• Both new assertions were confirmed to fail against the implementation each
  commit replaces.

Summary by CodeRabbit

• Bug Fixes

  • Improved reverse string searches for binary and single-byte strings by consistently using byte offsets.
  • Preserved character-aware reverse searching for standard UTF-8 strings.
  • Corrected handling of negative positions, bounds, empty searches, missing matches, and multibyte content.
  • Ensured String#byterindex correctly searches by byte position, including within multibyte UTF-8 sequences.

• Tests

  • Added coverage for byte-based reverse searches across UTF-8, binary, and ASCII strings.