LoongArch: Add ifunc support for memchr{aligned, lsx, lasx}

According to glibc memchr microbenchmark, this implementation could reduce
the runtime as following:

Name               Percent of runtime reduced
memchr-lasx        37%-83%
memchr-lsx         30%-66%
memchr-aligned     0%-15%