Remove need for SSE2 support

[satta]

Currently the code in ksw.c seems to need SSE2 to compile. It would be nice to have some kind of -- probably slower -- fallback implementation to improve support on architectures without these instructions.

[lh3]

I need to revamp ksw.c at some point. It is in fact buggy in rare conner cases. Will add non-SSE implementation when that day comes.

[satta]

Thanks!