Support report-only criteria without discarding their declared weight #101

Problem

A criterion can currently be excluded from task pass/fail and weighted scoring only by setting weight to 0. Some evaluation contracts need to preserve a non-zero declared weight for denominator/provenance parity while making a criterion genuinely report-only.

Using pass_threshold: 0.0 keeps a binary run_command miss from failing final_status, but calculate_weighted_score still includes its score and weight. The persisted CriterionResult also reports gating: true because gating is derived solely from weight > 0.

Reproduction

With two run_command criteria:

• criterion A: score 1, weight 8, pass_threshold 1
• criterion B: score 0, weight 1, pass_threshold 0

The task finishes SUCCESS, but weighted_score is 8/9 and criterion B is persisted as gating: true.

Requested contract

Add an explicit scoring/gating control, or document and implement pass_threshold: 0.0 as report-only, so a checker can still run and record its raw outcome while being excluded from both final_status and weighted_score without rewriting its declared weight.

The HTML/JUnit pass counters and persisted gating field should follow the same source of truth. A regression test should cover a failed report-only criterion with a non-zero declared weight.