Use Python's unicodedata.name() in fontTools.unicode

This removes our internal copy of Unicode names database!

Adds new API fontTools.unicode.setUnicodeData() that can be
called with a filename or file object corresponding to a
UnicodeData.txt file, which will consequently used for Unicode
character name mapping.  This is useful for using newer Unicode
database than the one that comes with the builtin Python module.

This also changes behavior such that control characters, Hangul
syllables, Han chars, etc, get no name with custom
UnicodeData.txt.  We may revisit this though.

Filed https://github.com/behdad/fonttools/issues/82 to add
option to ttx to pass custom UnicodeData.txt.

Fixes https://github.com/behdad/fonttools/issues/81