search:handle ^ as first character in character class

Created by: stefanhengl

Relates to #12476

Previously we did not handle ^ within character classes so that [^-b] (glob) was translated to ^[^-b]$ (regex) which is not the same.

Additionally, we handled - very strictly, following the UNIX man page for glob. However, we can be much more forgiving because google re2 can handle even those cases we previously excluded, like [a-b-c] or [a-z--0].

Because of the code change, a switch block in translateCharacterClass became obsolete and I removed it, which unfortunately bloats the diff. However, I didn't change the tests apart from adding new ones and moving [a-b-c] or [a-z--0] from "not allowed" to "allowed".

Note: I fuzzed the parser without any crashers.