Copy encoding when escaping hrefs/HTML

[kivikakk]

Fixes #130.

Looking at this code makes me feel weird. I guess we kind of leak result in these methods. They're allocated from the arena the node belongs to, so they get reset whenever rb_markdown_to_html is called, which should be often enough, but if people don't use that method, these might be leaky.

[gjtorikian]

If you feel up to it, I'd probably argue for some kind of c_str_to_rb_enc helper method to wrap these up, but eh, this is fine. I made decent progress on the Rust port so I don't think thinking much longer on the C maintainability is worth it.

[kivikakk]

Good thinking. I've named it to fit the other encoding helper. (I wonder if I shouldn't just use that here, because it's unlikely non-UTF-8 data could even be used with this.. but leaning toward not thinking much longer on C maintenance.)