fix(daily-syntax-error-quality): reduce token usage from 6M to ~150-400K/run (#29169)

* Initial plan

* feat: optimize Daily Syntax Error Quality Check to reduce token consumption

- Pre-select 5 random candidates in pre-step (was 235 candidates listed)
- Remove cat .github/workflows/*.md bash permission (prevented mass file reads)
- Remove open-ended find and head -n * bash permissions
- Fix turn target: 30 -> 20 (corrects accidental regression from 2026-04-06)
- Lower early-stop threshold: >=70 -> >=65 (fires more often for healthy runs)
- Lower individual score floor: <55 -> <50
- Condense verbose scoring rubric ~115 lines -> compact table ~15 lines
- Update scratchpad/token-budget-guidelines.md with new revision history

Expected token reduction: 6M -> 150K-400K per typical run

Agent-Logs-Url: https://github.com/github/gh-aw/sessions/ff17a14e-a4cf-4c57-96ab-3ce4329bce12

Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: gh-aw-bot <259018956+gh-aw-bot@users.noreply.github.com>

Author: Copilot