Allow &[u8] and Vec<u8> input

[tomekb234]

This library immediately converts &str input to &[u8] with as_bytes() and then does not appear to assume that the input is UTF-8 encoded, in the sense that non-UTF-8 input does not cause any undefined behavior, panics, or an invalid generated DOM. Since there is no parse() variant with &[u8] input, given such an input to parse, it is necessary to awkwardly use the unsafe str::from_utf8_unchecked to avoid unnecessary UTF-8 check overhead.

This pull request:

• changes the type of the input parameter of Parser::new from &str to &[u8],
• changes the type of the input parameter of VDomGuard::new from String to Vec<u8> (note that as String is backed with Vec<u8>, conversion from the former to the latter is cheap),
• changes the type of the pointer stored in VDomGuard from *mut str to *mut [u8],
• adds new public parse_bytes() and parse_bytes_owned() functions,
• adds a test for non-UTF-8 input.

Fixes #61.