What the tokens bought, by pillar and by milestone: a limb trimmed on evidence #1067

Replaces #828, parked twice:

• the check failed ("capacity: only a forge admin reads or changes it > an admin changes each one from the page, the API and MCP; a wrong value is refused; every change is kept") and the decider saw no point in another try (pull #1017 closed). Planner: split this smaller, or rewrite what done looks like; a person can also say what to try.
• the check failed ("every open pull's preview on the studio page > a change's page names its preview's data as its own copy and gives the moment it was taken; production's card does not") and the decider saw no point in another try (pull #1040 closed).

Those checks went red on tests this change does not touch. This waits on #1047, which takes them out of the gate, so the check measures this change. Start from today's main. What done looks like is #828's, unchanged:

---

Pillar: P7 Steering. The second part of #798, which #811 closes. One run, after #708 lands.

Why. The owner: "we need to be sure that what we are spending tokens on is valuable… I want to get to a point where we are capturing token ROI." #811 says what the week's tokens bought per change and what landed nothing. It cannot yet say which part of the vision the money went to, or let a limb of the map be trimmed on evidence.

The job: the same spend-against-landed figures, by pillar and by milestone.

Acceptance.

• [ ] tokenReturns (src/core/returns.ts) also groups by pillar (#708's pillar on each issue): dollars, changes landed, dollars per landed change, closed with nothing, still open.
• [ ] A milestone's page shows what its limb has cost and landed, with its costliest issues, so a limb can be trimmed on evidence.
• [ ] /board/returns gains a "By pillar" card; GET /api/v1/returns and the MCP returns carry the same figures. Owner facts only, as in #811.
• [ ] Tests: a fixture across two pillars adds up; a reader and an anonymous visitor get none of it.
• [ ] Reachable: the page, the API, the MCP, looks at phone and desktop widths.

Must not touch: the release workflow, the chart, other repositories.

Parked by the crew: the check failed and the decider saw no point in another try (pull #1232 closed). Planner: split this smaller, or rewrite what done looks like; a person can also say what to try.

Parked by the crew: the check failed ("a full queue shows in queue, and an old one in needs") and the decider saw no point in another try (pull #1513 closed).

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

1. the check failed and the decider saw no point in another try (pull #1232 closed)
2. the check failed ("a full queue shows in queue, and an old one in needs") and the decider saw no point in another try (pull #1513 closed)

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.

Split into #1594 (tokenReturns by pillar on /board/returns), #1595 (the API and MCP) and #1596 (a milestone's cost and landed). What done looks like is unchanged. Each piece is smaller.