Recover Rip access logs quietly after Janus restarts

[shreeve]

Restarting Janus drops its in-memory app registrations. Rip's access-log client retried the deleted IDs and printed every failed attempt, producing repeated socket errors and HTTP 404 diagnostics until the manager re-registered.

Stop retrying an ID after a 404 and wait for the manager to move the stream to its replacement registration. Report a repeated connection error once until a valid hello resets the diagnostic, while still reporting changed errors. Access streams and manager heartbeats share “Janus unavailable” diagnostics for known connection failures instead of Bun's misleading URL/port typo suggestion.

Validation:

• Regression tests reproduce repeated dead-ID requests and duplicate diagnostics before the fix.
• All 139 Rip Sites tests pass, including a real Janus restart held offline through a heartbeat, followed by app re-registration and resumed access output.
• Language suite: 3,196 passed.
• Browser suite: 16 passed, 2 existing skips.
• Local app managers have been restarted to load the fix; a live Janus restart recovered all four apps.