Cut a bunch of unused code out of mbstring

[alexdowad]

The legacy mbfl_strcut function is only used to implement mb_strcut
for legacy text encodings which 1) do not use a fixed number of bytes
per codepoint, 2) do not have an 'mblen_table' which can be used to
quickly determine the codepoint length of a byte sequence, and 3) do
not have a specialized 'mb_cut' function which implements mb_strcut
for that text encoding.

Remove unused code from mbfl_strcut, and leave only what is currently
needed for the implementation of mb_strcut.

---

Over the last few years, I refactored mbstring to perform encoding conversion
a buffer at a time, rather than a single byte at a time. This resulted in a
huge performance increase.

After the refactoring, the old "byte-at-a-time" code was retained for two
reasons:

1. It was used by the mailparse PECL extension.
2. It was used to implement mb_strcut for some text encodings.

However, after reviewing mailparse's use of mbstring, it is clear that
mailparse only relies on mbstring for decoding of QPrint, and possibly
Base64. It does not use the byte-at-a-time conversion code for any
other encoding.

Further, mb_strcut only relies on the byte-at-a-time conversion code
for a limited number of legacy text encodings, such as ISO-2022-JP,
HZ, UTF-7, etc.

Hence, we can remove over 5000 lines of unused code without breaking
anything. This will help to reduce binary size, and make the mbstring
codebase easier to navigate for new contributors.

[alexdowad]

@remicollet I ran the tests for mailparse against this change to ensure it doesn't break anything. Tests for mailparse are still passing.