Parsers without error recovery should fail faster on invalid inputs

[bd82]

Currently a re-sync recovery can still be triggered at the top most rule.
This will always resync to EOF which is the "last" token.

However this means iterating over all the Tokens which takes a while.
In such a scenario it is better to just "hop" to the end of the token vector immediately.

perhaps this is relevant even for "recovery enabled" parsers. as EOF should only appear at the end of the input...