Behavior of rules in character sets is underspecified

[codesections]

Roast has many tests where a rule is combined into a character set. For example, S05-metasyntax/charset.t. However, all of the current tests use a rule that can match against a single character. This means that the behavior of a token that requires multiple characters is unspecified.

I believe that the correct behavior is, to quote jnthn from a different issue:

<+foo> if foo matches more than one char [means that]: it has to match, but we only advance the cursor by 1 position however much it matches.

Unless anyone disagrees with this behavior, it should be specced, probably in S05-metasyntax/charset.t.

[codesections]

Cross references:

• Clean interpolation for enumerated character classes problem-solving#97
• Rules in character sets incorrectly consume multiple characters rakudo/rakudo#5430
• Rules in character sets undocumented  doc#4389