Implement dumb Unicode string search, and eliminate old implementation.

Implement dumb Unicode string search, and eliminate old implementation.

It looks like the KMP string-search algorithm is better for Unicode than BM, so
that will need to be implemented soon-ish.