Feature: Character Encoding Auto Detection

In v0.0.21, check-spelling assumes ASCII encoding and roughly ignores anything outside [a-zA-Z'].

In order to support non-English languages, the acceptable character-set will have to grow to include:

• Spanish (plus uppercase): á é í ó ú ü ñ ¿ ¡
• French (plus uppercase): é à è ù â ê î ô û ä ë ü ç

... And most other languages add other characters. Sadly, these characters are outside basic ASCII and close enough to having a conflict between Latin-1 and Unicode.

It would be nice to support latin-1, utf-*/ucs-*, and possibly some alternate encodings.

One approach would be t recognize some sentinel bytes, and once those are present, switch to a mode where the file might be worth checking again in an alternate encoding. In that mode, if the file has a words vs non-words comparison that is close to the normal threshold, it can be re-checked w/ another encoding, and if that pushes it into the right category, it can be added w/ a hint to add it to a pattern mapping to the selected encoding.

This would require adding a second threshold for truly non-word files (currently a single threshold exists for binary files where non-words exceeds words).