Health check should support configurable consecutive failure/success thresholds before marking a node as error or recovered

[mustfqyl]

Add configurable node health check thresholds to reduce false positives.

Currently, a single failed health check can mark a node as error or trigger a reconnect. In real-world environments, temporary network issues, API delays, or short Xray restarts may cause false downtime reports.

Suggested settings:

• NODE_HEALTH_FAIL_THRESHOLD=3
• NODE_HEALTH_RECOVER_THRESHOLD=2

Example behavior:

• Mark node as error only after 3 consecutive failed checks.
• Mark node as healthy only after 2 consecutive successful checks.

This would significantly improve uptime accuracy and prevent unnecessary reconnects.