Reduce and document Daily Observability ET budget (80M → 20M)

[Copilot]

daily-observability-report.md carried an outlier max-effective-tokens: 80000000 without auditable rationale, despite this being a daily workflow. This change right-sizes the cap to observed usage and records the evidence inline so future audits can evaluate it quickly.

• Budget calibration

  • Lowered max-effective-tokens from 80000000 to 20000000 in daily-observability-report.md.
  • Keeps substantial headroom over recent successful run usage while removing the extreme ceiling.

• Auditability in frontmatter

  • Replaced generic comments with concrete run-based evidence (run IDs + observed ET values).
  • Encodes the rationale where the budget is configured, reducing future ambiguity.

• Compiled workflow parity

  • Updated generated lock output so apiProxy.maxEffectiveTokens matches the new source budget.

# Tuned from 80M after checking recent successful runs:
# - run 26482328945: 2,674,398 ET
# - run 26424799100: 7,479,570 ET
# - run 26197188358: 7,244,916 ET
# 20M preserves >2.5x headroom over the recent peak while keeping daily budget hygiene.
max-effective-tokens: 20000000