currently a collation lookup (collation name to collation id) is done (in get_collation_number_internal()) by linearly scanning the array of collations.

This used to work fine, but the number of collations is constantly growing (MDEV-27009, MDEV-20912) and the linear search is starting to become noticeable.

It would be a good idea to change it to a hash lookup.

Same can be done for character sets, even though the number of charsets didn't grow that much.