docs: design doc for a built-in Tiny system-health DM

[graycyrus]

Closes #2375

Summary

Design doc for a permanent, non-removable "Tiny" DM that surfaces system
health (connections, budget, workflow failures, approval backlog) and, later
with approval, can act on some of it.

• docs/issue/tiny-agent/README.md — overview and reading order
• docs/issue/tiny-agent/design.md — the core decision (identity separate
  from cognition: a reserved id + pinned DM, not a roster [[agent]]), the
  trigger-loop architecture, and a per-signal cost table (connections/
  credits/usage), each claim anchored to the current source it's based on
• docs/issue/tiny-agent/edge-cases.md — edge cases and explicitly open
  decisions (relationship to NotificationStore/EPIC: approvals interrupt for things that carry no consequence, and every interruption dead-ends the turn #558, what "credits" means,
  rename/rollout policy)

No runtime code changes. This is the spec a follow-up implementation issue
will scope Phase 1 work items from.

Test plan

• docs/issue/tiny-agent/*.md each under the repo's 500-line cap
  (wc -l: 29 / 210 / 79)
• Every file:line citation in design.md checked against current
  main (RESERVED_AGENT_IDS, SYSTEM_AUTHOR, CONFINED_AGENT_ID,
  WORKFLOW_REPLY_AUTHOR, MaintenanceTicker, BudgetPauseMarker,
  project_connections, remove_member) rather than carried over
  unverified from research
• Docs-only change; no cargo fmt/clippy/test needed

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

_{Comment @coderabbitai help to get the list of available commands.}