sequences. Moved the check for overlong sequences, not it's done
after the complete sequence is processed.

* src/charset.c (low_utf8_to_local): If utf8_tolerant is set,
replace invalid utf8 sequences with '?'. Used mainly by the
testsuite.

* src/charset.h (enum utf8_flag): New flag utf8_tolerant.

Rev: src/charset.c:1.13
Rev: src/charset.h:1.9
Rev: src/parse.c:1.43