Fix mutations Stryker surfaces

[Mearman]

Once stryker.config.ts exists and a baseline run completes, go through every surviving mutant: either the test suite is missing an assertion that should have caught it, or the mutated code path is genuinely dead and can be deleted. Don't chase 100% for its own sake where a mutant is equivalent, but a surviving mutant on a real branch (a swapped comparison in peer-id verification, an off-by-one in stale-agent cleanup timing, a dropped early return in delivery-status handling) is a real gap and gets a real test.

Target 80% mutation score on every file stryker.config.ts's mutate list currently names.

Scope changed after #97 split core/mesh-store.ts into ten collaborator files (agent-registry.ts, connection-approval.ts, delivery-engine.ts, federation-bridge.ts, mesh-store-shared.ts, peer-lifecycle.ts, room-lifecycle.ts, room-messaging.ts, room-protocol.ts, room-wire-extensions.ts), with mesh-store.ts itself reduced to a thin orchestrator. The branch-heavy, security-relevant logic this issue originally targeted in mesh-store.ts now lives across those files, so #98 expanded the mutation-testing CI matrix and stryker.config.ts's own mutate list to cover each of them individually. This Epic now tracks one Task per file, rather than the original two-file list.

Current status (measured against the real, current split):

File	Score	Status
identity.ts	92.86%	Done (PR #100)
wire-mesh-transport.ts	78.70%	PR #101 merged, remaining gap tracked in #114
room-wire-extensions.ts	100.00%	Done, no work needed
agent-registry.ts	76.19%	#104
room-protocol.ts	69.57%	#111
room-lifecycle.ts	65.20%	#109
peer-lifecycle.ts	64.52%	#108
mesh-store-shared.ts	66.67%	#113
mesh-store.ts	60.19%	#103
connection-approval.ts	57.14%	#105
delivery-engine.ts	56.64%	#106
room-messaging.ts	55.71%	#110
federation-bridge.ts	49.02%	#107
stale-agent-checker.ts	45.65%	#112