memchr(3) rewrite
I have taken some time to rewrite memchr(3) to utilize a similar
algorithm to my strlen(3). Two different formulas, however, would
have roughly a 1/3 difference.
Currently I have taken an approach where, when memchr(3) tries to
match an 8-bit pattern, it treats the string as 8-bit using the first
formula, which is expected to have uniform time complexity by
reducing unnecessary branches.
This approach, however, might be bogus: if you look for a 7-bit
character within an 8-bit string, memchr(3) will perform poorly when
there are many characters with their MSB set to 1.