Improve the performance of Unicode Normalization Forms.

This changes significantly speeds up the normalize() function.
Similar to the changes in case mapping (fold, lower, upper, or title case),
here we get rid of perfect hashing and replace it with a Sparse Array with
a nested branch function to check specific ranges of code points.
This allows us to achieve a performance increase of up to 40%.

Author: Alexander Borisov <[email protected]>

Author: lexborisov