The information if a validator outputs `score.txt`, `score_multiplier.txt`, or neither should be available statically

[gkreitz]

A validator can output score.txt, score_multiplier.txt, or nothing. It must pick between score.txt and score_multiplier.txt across invocations per https://www.kattis.com/problem-package-format/spec/2025-09.html#reporting-additional-feedback :

The validator may create a score.txt, a score_multiplier.txt, or neither, but it must not create both. This applies over all invocations. That is, if a validator ever creates a score.txt, then it must never create a score_multiplier.txt, and vice versa.

There are two problems here. Firstly, to figure out if a validator produces score.txt or score_multiplier.txt, we must run it. Secondly, if I read the spec correctly, a validator can always decide to output neither file (which will be interpreted as "full score"). This feels like a fragile design if there are any buggy paths in a validator causing it to for some reason fail to write a score. (Also, as a remark, this second problem also means we can't even run the validator once and see what it produces, in the general sense, figuring out if a validator produces score.txt or score_multiplier.txt is likely equivalent to solving the halting problem. :))

For legacy problems, we require an output validator for a scoring problem to always output a score. I think the same should hold for the new version - a problem author should specify somewhere which of the 3 modes a validator uses to emit a score (score.txt, score_multiplier.txt, or neither). I would strongly prefer this information to be available without having to compile and run the validator.