The hand-written `to_upper`/`to_lower`/`to_title` tables only covered a small
part of Unicode and contained wrong entries, and `to_title` searched
`to_upper_singlets` using the length of `to_title_singlets`.
The UCD generator now reads the simple case mappings from UnicodeData.txt and
emits the tables into `generated.odin`. The titlecase tables only hold the
runes whose titlecase differs from their uppercase, with `to_title` falling
back to `to_upper` for the rest.
Fixes#7910
Matthew Winter
5f16eb2832