Autocorrect should not run on files where ParserError fails

[TastyPi]

I had a file that failed the ParserError check while running with autocorrect, and another lint removed some code in the file after the parser error. I didn't realise the code was removed until submitting the code for review and checking the diff.

I then went back to what I had before running erblint, fixed the ParserError first, then ran autocomplete, and it worked correctly. It would be nice if erblint checked ParserError first and, if that fails, don't bother with the other lints as they're likely to run into parsing errors too. At the very least it shouldn't autocomplete anything.