Add a more advanced normalize function

[bact]

Common normalize function that do more than reordering of Thai characters. Something that can be used quickly for matching, searching, sorting, and preparing data for classification tasks.

Some ideas;

• Remove non-visible characters, like zero-width chars Add a function to remove zero-width characters #373
• Remove unnecessary spaces
• Normalize repetitions
• Normalize "obvious" mistakes like
  • consonant + tonemark A + tonemark B <--- may be we can remove tonemark A

Note that for Unicode normalization, Python does already have unicodedata.normalize().

[p16i]

How about the case of ำ? I sometimes see PDF parsers return it as ◌̊ and า.

[bact]

Current normalize() will group Nikhahit and Sara Aa together and emit Sara Am