RIMAPI: expose AlertsReadout and LearningReadout via API

[CalebisGross]

Summary

RLE agents would benefit from access to RimWorld's built-in alert system (AlertsReadout) and learning helper (LearningReadout). These are the notifications shown in the top-right of the screen (e.g. "Major break risk", "Need colonist beds", "Need meal source", "2 colonists idle").

Proposed endpoints (in RIMAPI fork)

GET /api/v1/alerts

Return the active alerts from AlertsReadout:

{
  "data": [
    {"label": "Major break risk", "severity": "critical", "colonist_ids": [184]},
    {"label": "Need colonist beds", "severity": "minor"},
    {"label": "Need defenses", "severity": "minor"},
    {"label": "Need meal source", "severity": "minor"},
    {"label": "2 colonists idle", "severity": "minor", "colonist_ids": [181, 187]}
  ]
}

GET /api/v1/learning-helper (optional, lower priority)

Return the active learning helper concepts from LearningReadout.

Why

Some of this data can be derived from existing endpoints (mood from /colonists, beds from /map/rooms), but RimWorld's alert system already aggregates and prioritizes these signals. Exposing it directly would:

1. Give agents a pre-prioritized list of colony problems without needing to infer them
2. Surface alerts that aren't derivable from current endpoints (e.g. "colonists idle", "need defenses")
3. Reduce prompt tokens — a compact alert list replaces multi-endpoint inference logic

Context

Discovered while running RLE against a live colony. The alerts visible in the game UI contain actionable information that agents can't currently access through the API. Issues are disabled on the RIMAPI fork so tracking here.