Cost ledger: measure compute for deploys and agent runs, not only checks #1388

Follows #1386. Checks pass a CellUsage accumulator into their cell (apps/contain.ts meter) and finishRun records cpu_s/mem_gb_s. Deploys (install + test steps in apps/process.ts startApp, and the serving app between deploys) and agent runs (the crew sandbox, crew/sandbox.ts) still enter null compute. Thread the accumulator through runStep/startApp for deploys, and sample the agent process tree the same way for agent runs.

Tests: a deploy entry carries non-null cpu_s when cells are available (skipIf no bwrap); an agent run's entry does.

Parked by the crew: the run ended without a pull: declined: #1388 builds on #1386's work, which isn't in this repository. Nothing was pushed.

Parked by the crew: the run ended without a pull: declined: nothing has changed since my first decline, so I still can't do #1388. The only new message on the issue is the crew's note that my earlier run was parked. origin/main is still at 308286d0, and it

Parked twice since a person last spoke here, so it is the planner's now. Every reason:

1. the run ended without a pull: declined: #1388 builds on #1386's work, which isn't in this repository. Nothing was pushed
2. the run ended without a pull: declined: nothing has changed since my first decline, so I still can't do #1388. The only new message on the issue is the crew's note that my earlier run was parked. origin/main is still at 308286d0, and it

The planner splits it, replaces it with a fresh issue, or defers it; a person can also say what to try. Nothing starts here until then, and the board lists it.

Deferred by the planner. This follows #1386 and builds on its core/cost_ledger.ts and core/ledger_write.ts, which are not on main. The map had it backwards: #1386 waited on this issue, so a builder was handed it before its foundation existed. Twice. It now waits for #1386, and it sits behind the Later gate (#595) because #1386 serves no pillar yet. Lift both links once #1386 lands and the owner opens it.