Epic: issue-quality measurement — score a scoped issue by how cleanly it executed

[JSONbored]

Problem

We author scoped issues and contributors execute them, but nothing measures whether an
issue was well-formed. A good issue produces a clean first-attempt merge; a bad one
produces revision churn, rejection, abandonment, or a post-merge revert. That signal is
already latent in the data and is never computed, so issue-drafting quality can only be
asserted, never demonstrated or optimised.

Why it is computable today

decision_records (migrations/0179) holds one row per (PR, head sha) with the acted
action (merge | close | hold) and reason_code. Multiple rows for one PR therefore
encode revision depth directly. pull_request_reviews and check_summaries add review
and CI outcomes, and the existing github-truth outcome scan supplies merged / reopened /
reverted confirmation.

No new instrumentation is required — this is a query over history we already hold.

Sub-issues

• define the score (which signals, how combined, documented)
• compute and backfill it across all historical issues
• expose it as an internal surface for drafting feedback

Acceptance

Every historical scoped issue carries a quality score derived from how its PRs actually
executed, reproducible from decision_records by anyone with the published definition.

[linear-code]

JSO-7074