Some charset + collation combinations are only settable via `charsetNumber`

[mscdex]

Example: utf8mb4_uca1400_ai_ci (charset number 2304)

Currently you can only set it via charsetNumber because setting collation: 'uca1400_ai_ci' will pick up the first charset it finds (which is utf8mb3), even if there are multiple matches.

Specifying charset: 'utf8mb4', collation: 'uca1400_ai_ci' also does not work as collation will get ignored since charset is set. In this particular case, the 'unicode_ci' collation would end up getting chosen.

I don't know how stable charset numbers are, but it would be better if charset: 'utf8mb4', collation: 'uca1400_ai_ci' would work the way you'd expect as it's clearer.