Add support for the same encodings that ripgrep supports

[acheronfail]

We can't trust the absolute_offset that ripgrep reports for non UTF-8 encoded files (see https://github.com/BurntSushi/ripgrep/issues/1627#issuecomment-649858419). So we need to parse the file ourselves.

Goals for this issue:

• [x] Use the same approach to encoding sniffing that ripgrep uses, either:
  • checking for a UTF-8 or UTF-16 BOM, and then using that encoding (defaulting to UTF-8 otherwise)
  • using the encoding passed on the command line
• [x] Find the exact location of the match a non UTF-8 encoded file, and insert the replacement text in the specified encoding. We changed tactics, but the result is the same. We now decode into UTF8/ASCII, perform the replacements and then re-encode before writing to disk

Supported encodings (tests exist for them):

• [x] ASCII
• [x] UTF8
• [x] UTF16BE
• [x] UTF16LE
• [ ] TODO: get a list of all the encodings ripgrep supports (uses the encoding_rs crate)

[acheronfail]

I'm going to close this for now.

If you spot any specific encoding issues while using rgr, please create an issue!

[acheronfail]

We can improve the encoding situation for rgr by contributing an upstream change in rg, see: https://github.com/BurntSushi/ripgrep/issues/1629