streamsync: speed up stream rediscovery and recover faster from local outages

[GheisMohammadi]

This PR Improves stream sync recovery when peers drop or the node briefly loses network, so sync can restart without long stalls or a process restart.

What changed

• Lower mainnet peer floors so sync can start with fewer streams (MinPeers=4, boot InitStreams=5, hard/soft discovery caps aligned).
• Shorter recovery timers: discovery watchdog 30m (was 75m), max removal cooldown 15m (was 60m).
• Advertise timing: use MaxAdvertiseWaitTime (mainnet 15m) as the real cap between advertise cycles.
• Fix DHT skip race: wait for trusted streams to actually register before skipping DHT discovery.
• Mass-disconnect detection: if many streams die quickly, treat it as a local network outage, allow fast reconnect for connection-loss failures, then rediscover after a short holdoff.
• Keep punishing bad peers: protocol/bad-data removals still get full critical cooldowns, even during a local outage; local-outage mode is also rate-limited (once per 10m).

How this helps

• Faster recovery after brief internet drops or peer flaps, without waiting over an hour for watchdog/cooldowns.
• Better discovery by not skipping DHT too early and by advertising more often when needed.
• Safer stream failure handling by distinguishing connection-loss from malicious/buggy peer behavior.