Preparation to relax iterator rollback policy, which is also currently violated by not rolling back skipper almost everywhere, so this should fix postskip disabling.

Under the current iterator rollback policy rollbacks could be omitted in compound parser which don't advance iterator themselves because the called subparser is responsible for rollback on failure. In reality these rollbacks are undoing skipper work what renders as inconsistency in error handling position (`p` vs `p >> eps` or `repeat(1)[p]`).