Search of a singleton needle should use memchr

[Bodigrim]

Surprisingly, both Data.Text.Internal.Search and Data.Text.Lazy.Search have special cases for searching a single ASCII character, but neither uses memchr, even while the main codepath does, so it is readily accessible.

(See https://discourse.haskell.org/t/blog-5x-speedup-changing-break-to-elemindex/13946 for context)

[phadej]

I bet there is a simply historical reason: memchr weren't applicable when text was using UTF16. (And main codepath memchr seems to look for a potential position of a needle; that works using ~any encoding).