Make nibble2base faster using x86-64 pshufb instruction (SSSE3) and using dynamic dispatch.

[rhpvorderman]

See https://github.com/samtools/htslib/pull/1677 for prior discussion.

I made the PR such that nibble2base is dynamically dispatched on x86-64 cpus with SSSE3 instructions.

No build options need to be changed and nibble2base gets a nice speed up on the majority of the install base.