MAINT: Mark `npy_memchr` with `no_sanitize("alignment")` on clang

Clangs sanitizer reports unaligned access here, which is correct
but intentional.  It may well be that the code would be better of
trying to avoid this unaligned access (and rather vectorizing harder).

But, this is a bit of a tricky choice, since we have to optimize for
different use-cases (in particular very short scans may be interesting).

So changing this would best be done together with some more careful
benchmarks.

See also gh-21117, which introduced manual loop unrolling to avoid the
unaligned access.

Closes gh-21116

0 files changed, 0 insertions, 0 deletions