Function compute_correctness 

pub fn compute_correctness(check: &CheckResult, expected: &TaskExpected) -> f64

Compute raw correctness score from check results.

Returns a raw score in [0.0, 1.0]. Sub-weights are fixed: fmt=0.10, clippy=0.30, test=0.60. If a check is not required by the task (TaskExpected), full credit is given for that sub-weight regardless of the actual result. Task weights are NOT applied here — they are applied in compute_scores.