Implemented guard enforcement semantics for denylist/allowlist checks,
budget caps, tool-call limits, and write/apply approval gates with
clearer denial messages and remediation guidance.
Extended guard evaluation to include plan vs subplan scope via a
GuardScope(StrEnum) for type-safe scope handling, and propagated
scope through the automation profile service guard checks. Extracted
remediation guidance strings as module-level constants for consistency
and testability.
Added Behave coverage for remediation messaging and subplan-scoped
tool-call limit behavior.
ISSUES CLOSED: #853
brent.edwards