Audit-reframe docs: surface 'simulations auto-run hourly via cron' (follow-up to #3734)

[koala73]

Follow-up filed from PR for issue #3734 (simulation trigger endpoint + runId filter).

The trigger endpoint we just shipped works around the audit's literal complaint, but the audit's framing — 'agents are permanently read-only' — was never accurate to begin with: the deep-forecast cron has been auto-enqueuing simulation tasks every cycle via the call at scripts/seed-forecasts.mjs:16096. That story is currently only documented in code; the user-facing OpenAPI / Pro marketing copy / MCP tool descriptions never explain it. Next external review will reach the same wrong conclusion.

Scope:

• Update OpenAPI description on GetSimulationOutcome to note 'outcomes auto-refresh on the deep-forecast cron cadence'
• Update Pro marketing copy to surface the auto-trigger behavior (so users know why fresh simulations appear without action)
• If/when C3 ships the MCP tool wrapper for TriggerSimulation, the tool description must explain 'usually unnecessary — outcomes auto-run hourly'

Related: tracking issue #3798 (the plan for the C1+C2 work), #3726 (audit umbrella).