Support extended Latin characters in auto-completion list

[zufuliu]

Currently only _0-9A-Za-z and some punctuation (depends on current scheme) is included when populate auto-completion list, because other characters can't (or not recommended to) be used as an identifier in most programming languages.
However, this will truncate words in many languages. e.g. Café is truncated as Caf.

[zufuliu]

Implement this will require decode bytes retrieved from Scintilla to Unicode code point (UTF-32), and then checking character ranges.

For Unicode (UTF-8, UTF-16) encoding, decoding is simple.
For one byte extended ANSI (like Windows-1252) encoding, decoding need a map or a call to MultiByteToWideChar(), which is slow.
For DBCS/MBCS (CJK) encoding, decoding need a call to MultiByteToWideChar(), which is more slower than one byte extended ANSI.

In commit 7063 (0d5edc93e280), Scintilla introduced some new methods for indexing document, maybe useful to implement this.

[zufuliu]

Fixed by commit c4f533e and f449b46.

For DBCS (Chinese, Japanese, and Korean), see https://sourceforge.net/p/scintilla/feature-requests/1226/.