Declare layer machine-readably (§11); adopt v0.6 corrections

The standard moved v0.4 -> v0.6. All four findings from our v0.4 review
were adopted in v0.5, and v0.6 went further on two of them.

§11 now requires a machine-readable declaration — prose cannot
distinguish a declaration from a transcribed review. We had none.
Added layer.yaml (form adapted from ops-warden's reference
implementation), scripts/check_layer_conformance.py, and
tests/test_layer_conformance.py.

The check makes our central claim mechanical rather than asserted: no
direct Tooling client in src/. The test exercises the negative case on a
synthetic tree, so it fails if the checker goes blind. pyyaml is added as
a DEV dependency only — `dependencies = []` is load-bearing for the §5
claim and stays empty.

Adopted from v0.6:
- containment is no longer ours (§9.2). Actuation is an Engine concept,
  unowned and held at zero; kings-guard proposes containment and never
  performs it. The register row is now a dependency, not our gap.
- observation is scoped to Staff-reachable sources, with identity and
  secret observation pending — our finding 1, adopted near-verbatim.
- access-engine DECLINED the authentication-evidence gap; owner is now
  the identity layer plus audit-core, reproposed and unassented.
- §11 blocked-clean recorded, with the rule that it must not rank below
  conforming — our finding 2.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UEtvmYUBP2fDtirJGWn5MW

Assistant: claude-code
Assistant-Model: opus
Assistant-Process: 4014379@bnt-lap001
Assistant-Session: 4af9e20f-1768-4afc-951b-b507784e382b

SCOPE.md

@@ -9,7 +9,7 @@
 **Staff** — interactive, non-deterministic; adaptive defence, observation,
 containment. Binding rule: kings-guard never touches Tooling directly; it acts
 only through Engine APIs. See `INTENT.md` and
-`net-kingdom/canon/standards/security-layer-model_v0.4.md`.
+`net-kingdom/canon/standards/security-layer-model_v0.6.md`.
 
 ---