lib/ext2fs: port fast crc32c routines for Intel CPUs from kernel.

[pdlan]

This PR ports kernel's fast implementation of crc32c to libext2fs. A simple benchmark by using fio to measure the write throughput of the fuse2fs FUSE driver on an ext4 partition (with 1M block size) shows that the throughput increases from 186 MiB/s to 278 MiB/s with my Micron SSD.

[tytso]

This patch has significant portability problems. It appears to assume that certain header files like /usr/include/cpuid.h are always present. If you enable github actions, you'll see that your commit breaks quite a number of builds: https://github.com/tytso/e2fsprogs/actions/runs/8745541253