Add retention governance for v0.43.0

[crypt0rr]

What changed

• add an administrator-facing retention dry run and explicitly confirmed cleanup action for bounded operational state;
• retain notification events, deliveries, inbox state, blocked work, and delivery-attempt audit records indefinitely for existing accounts;
• make the existing maintenance windows configurable through an in-memory retention policy while keeping conservative defaults;
• align the Docker and CI toolchain on the pinned patched Go 1.25.6 release and validate that drift in make tooling-check;
• update documentation and tests for the new retention governance behavior.

Why

Operational records previously had maintenance behavior but no visible policy or safe operator workflow. This makes the eligible state inspectable and cleanup deliberate without silently removing notification history.

Impact

Administrators can review candidate counts in the Admin interface and must explicitly confirm cleanup. Existing event and delivery history is preserved; account deletion semantics are unchanged. Runtime/provider behavior and routes are otherwise unchanged.

Validation

• gofmt
• go test ./... -count=1
• go vet ./...
• go test -race ./... -count=1
• make coverage
• make lint
• make vuln
• make tooling-check
• docker build --target test .
• docker build -t artist-trackarr:v0.43.0 .
• scripts/container-smoke.sh artist-trackarr:v0.43.0