Fixes needed for scanning of large files

[tomlea]

Fixes to allow my use case of parsing 500mb+ file, with huge blocks of content that do not need parsing, but was taking forever to do without.

• Reparse for error logging no longer resets to 0, it returns to the starting position. This is helpful for parsing lots of entries from a stream.
• Adds Source#consume_until which allows me to scan large blocks of code looking for an end symbol.

[kschiess]

Sounds like a bit of a new use case for parslet; new in the sense that it was never going to work on huge inputs and wasn't made for it - parsing is just too slow. In your case: Would it be impossible to extract parseable bits by other means and then parse the remaining important parts with parslet?

If I merge and then announce parslet works with 500's of MBs, I am putting us in a bad spot...