Improve ACP UX when backend is ready but dispatch policy is disabled (`acp.dispatch.enabled=false`)

[kimptoc]

Summary (issue provided by Codex)

ACP can appear “broken” even when the ACPX runtime is healthy, because dispatch is separately gated by policy.

Current behavior

• Logs may show ACPX backend is ready (acpx runtime backend ready)
• But ACP turns fail with:
  • ACP dispatch is disabled by policy (acp.dispatch.enabled=false)
  • error code: ACP_DISPATCH_DISABLED

This is correct behavior, but the distinction is easy to miss during setup/troubleshooting.

Expected behavior

Make the dual-gate model explicit in CLI/log UX:

1. ACP backend availability (runtime/plugin)
2. ACP dispatch policy (acp.dispatch.enabled)

Suggested improvements

• On startup, if ACP backend is ready but dispatch disabled, emit a clear INFO/WARN like:
  • “ACP runtime is ready, but ACP turn dispatch is disabled by policy (acp.dispatch.enabled=false).”
• In ACP-related doctor/status output, show both states separately:
  • runtime: ready|not-ready
  • dispatch policy: enabled|disabled
• In docs/setup examples, include minimal required config together:
  • acp.enabled=true
  • acp.dispatch.enabled=true
  • acp.backend=<backend-id>

Why this helps

Reduces false diagnosis of “runtime unavailable” and shortens troubleshooting for new ACP users.