Implement a "token filter" layer or equivalent functionality for e.g. discarding whitespace

[pczarn]

Example

grammar! {
    // ...
    sub filter! {
        // ...
        discard: whitespace;
        sub grammar! {
            whitespace ::= r"\s+";
            // ...
        }
    }
}

Perhaps sub-grammars could automatically wrap themselves in token filters. This would reduce annoying nesting.