Show reasoning effort next to the model in the subagent progress line

Summary

The subagent progress line renders the model a subagent actually ran on, but not its reasoning effort. The main session's statusline shows both. Please show both for subagents too.

# today
ce:hand(worker/smoke:valid) Haiku 4.5

# asked for
ce:hand(worker/smoke:valid) Haiku 4.5 med

The same line is what the running-agent footer prints while a dispatch is in flight, so it is the only place a live subagent's tier is visible at all.

Why it matters

Effort is a real cost and quality axis. A dispatch at haiku/low and one at haiku/xhigh are very different runs, and today they render identically. When several subagents are in flight, there is no way to tell from the screen which ones are running expensively.

It matters more for any agent whose tier is chosen per dispatch rather than pinned in its definition. I maintain one (ce:hand — it adopts another agent's contract and runs it at a tier the dispatch names, so its frontmatter deliberately pins neither model: nor effort:). For that shape, what the dispatch declared and what the harness actually ran can diverge silently, and the progress line is where a human would catch it. It shows the actual model, so a model mismatch is catchable. Effort is not shown, so an effort mismatch is not.

Current workaround, and why it is not enough

The only text a dispatch controls on that line is the description string inside the parentheses, so both tiers get stuffed in there:

ce:hand(haiku/med worker/smoke:valid) Haiku 4.5

That works, but:

• It duplicates the model, which the harness already prints — noise on every line.
• It is self-reported. The tags say what the dispatcher believed; only the harness knows what actually ran. The value of the model suffix is precisely that it is not self-reported, and effort has no equivalent.
• It costs label width that would otherwise describe the task.

Related

• #43083 — configurable reasoning effort level for subagents (the control knob). This issue is about the display, and is useful independently: even with effort resolved purely from frontmatter or the session, showing it would remove the guesswork.
• #39220, #25669, #25591 — per-dispatch effort requests.

Notes

Rendering the resolved effort — whatever it resolved from (frontmatter, session, env) — would fully cover this. A -- or omission when it cannot be determined is fine; an honest blank beats a guess.