Implement Unicode TR-29 and TR-14

[rogerbinns]

This is needed to get #473 to work reasonably well, especially when one user perceived character is many unicode codepoints. The word splitting is also a lot better than trying to do it by codepoint category as in existing unicode61 (eg that messes up don't) and sentence splitting is good for a better snippet function.

• [x] Generate break tables
• [x] Test code from tables
• [x] Grapheme cluster
• [x] Word
• [x] Sentence
• [x] Unicode categories\
• [x] Equivalent_Unified_Ideograph
• [x] Emoji
• [x] Regional Indicator
• [x] Case folding
• [x] Investigate tr14 line breaking for implementing textwrap
• [x] textwrap
• [x] Wide codepoints (east asian width == F or W)
• [x] Grapheme cluster count
• [x] Grapheme cluster range substring
• [x] Grapheme cluster width
• [x] Startswith / endswith
• [x] find/index
• [x] Set doc order to bysource and rearrange functions into logical doc order
• [x] Grapheme cluster base char (remove diacritics equivalent)
• [x] Compatibility codepoints (eg roman numeral ⅲ becomes latin iii)
• [x] Update apsw.ext.format_query_table to use textwrap
• [x] Update tests
• [x] Convert important bits to C
• [x] Convert other bits to C
• [x] Rename module to unicode

[rogerbinns]

Equivalent_Unified_Ideograph should be investigated. Looks useful for stripped function for getting compatibility codepoint.