Docs: document rlmMaxDepth configurability (settings key + /rlm-max-depth)

[abhishakenp]

Summary

RLM recursion depth is fully configurable in the code, but this is undocumented:

• The rlmMaxDepth settings key exists at packages/coding-agent/src/core/settings-manager.ts:136
• The /rlm-max-depth [<int> [--global]] slash command exists at packages/coding-agent/src/core/slash-commands.ts:175
• The RLM_MAX_DEPTH env fallback exists in _resolveRlmMaxDepth at packages/coding-agent/src/core/agent-session.ts

However, neither docs/settings.md nor docs/rlm.md mentions any of these. docs/rlm.md only says "Raising the configured depth allows descendants to recurse further" without naming the knob, and docs/settings.md has no Recursion section. Users cannot discover how to raise the default depth of 1.

Verified behavior

We verified live in a multi-agent fleet:

• Setting rlmMaxDepth: 10 in user-level settings.json unlocked 10 levels of rlm(...) recursion
• /rlm-max-depth reported the effective depth and source (RLM max depth: 10 (global))
• The depth guard fired correctly once exhausted: RLM recursion depth limit reached (RLM_DEPTH=n, RLM_MAX_DEPTH=10)
• Resolution order in _resolveRlmMaxDepth(): persisted per-chat state → inherited parent value → settings.json → RLM_MAX_DEPTH env → default 1
• Existing tests covering the setting are unaffected (agent-session-recursion.test.ts, daemon-mode.test.ts)

Proposed change

Docs-only addition (+15/-1), verified against source:

1. packages/coding-agent/docs/settings.md — new "Recursion" section documenting rlmMaxDepth (type, default 1, 0 disables recursion), a JSON example, and the full precedence chain.
2. packages/coding-agent/docs/rlm.md — expand the one-line depth mention to name rlmMaxDepth, /rlm-max-depth, the precedence chain, and link to the Settings section.

A complete patch is ready on a public fork branch:
https://github.com/abhishakenp/prime-agent/tree/docs/rlm-max-depth

Happy to open a PR against main if maintainers invite/vouch — just say the word.