Do not generalize the alphabet, support only string and u8 slices

[orium]

We should also allow case-insensitive matching at both ascii and utf-8 characters level.

See this comment on reddit for more context.