v1.108.69 - durable-change delivery (get_delivery_metrics / delivery CLI)

Durable-change delivery: an honest output measure to pair with the token meters.

The suite shows tokens used and tokens saved (input). It had no honest measure of how much got done for that spend, the gap that lets teams reward raw volume over outcomes. get_delivery_metrics (MCP tool) and delivery (CLI) mine local git over a window and classify each non-merge commit into one mutually-exclusive bucket by precedence:

• revert_authored - the commit undoes earlier work
• reverted - a later commit reverts this one
• reworked - a file it touched was re-touched by a later commit within rework_horizon_days (churn-back)
• durable - landed and stuck

commits_durable is the honest numerator for a cost-per-outcome ratio: divide AI spend over the same window by it. The delivery CLI's --cost does exactly that and prints the headline cost-per-durable-change.

Honesty fixes baked in:

• Hub files co-touched by an outsized share of window commits (CHANGELOG, version files, a monolithic dispatch module) are shared ledgers, not rework evidence, so they're excluded from the rework signal (93% to 38% on this repo's own history; auditable via _meta.hub_files_excluded).
• Durability is trailing: commits inside the horizon are surfaced as commits_provisional (counted durable, not yet settled). Diagnostic trend, not a score to chase.

Standard tier, so core_compact is unaffected. Tool count 84 to 85 (full). Read-only git archaeology, no new tables. New tests/test_delivery_metrics.py (7). Full suite 4774 passed / 8 skipped; CI green across all 8 matrix jobs.