support character encodings in supported documents

[rhdunn]

The MIME headers (mhtml, email, http) can specify a character encoding. This should be adhered to when processing the internal data content (e.g. passed to the XML reader API).

Also, text documents can have a Byte-Order Mark that specifies what flavour of UTF encoding they are.

RTF documents handle their own encoding for escaped characters.

With text documents, it would be useful to specifically override/configure the default encoding (default as utf-8, override with e.g. cainteoir --encoding=latin-1 or as a preference/setting in cainteoir-gtk).

[rhdunn]

In addition to this work, the encoding APIs need improving:

1. refactor set_encoding(int codepage) to only call set_encoding(string encoding) once;
2. provide access to the current encoding name/codepage;
3. add tests for the encoding APIs (aside from what is tested by the usage of the encoding API);
4. don't reset the encoder if changing to the same encoding;
5. don't pass utf-8 encoded strings through iconv -- return the buffer directly (utf-8 => utf-8 should be a no-op);
6. provide some mechanism to signal to things like the xml reader that it is safe to parse the original buffer without decoding it all (utf-8, latin-1 or another single/multi-byte encoding with an ascii base) or whether it needs to decode the entire buffer first (minus any byte-order mark, for utf-16/32 le/be and any strange codepages that don't have an ascii base).

[rhdunn]

The encoding API work (with tests) is now in place.

[rhdunn]

This is done except for BOM handling.