fix(feishu): add WebSocket pingTimeout and lifecycle event logging

[yyteen-bot]

The Feishu adapter's WSClient was created without a pingTimeout, causing the Lark SDK to never detect half-open WebSocket connections. This led to bots becoming unresponsive after idle periods (several hours) without any error logs.

The Lark SDK's pingTimeout defaults to 0 (disabled), meaning the pong watchdog is never armed. When a WebSocket connection silently drops (e.g., due to NAT timeout, firewall, or WSL network sleep), the SDK never detects it and never triggers reconnection.

Changes:

• Add pingTimeout config option (default: 30s) to enable dead connection detection
• Add WSClient lifecycle event callbacks for visibility into connection state
• Fix stop() to properly close WebSocket before releasing references
• Expose pingTimeout as configurable via golem.yaml

Fixes #28

What does this PR do?

Fixes Feishu bot becoming unresponsive after idle periods by enabling the Lark SDK's WebSocket pong watchdog (pingTimeout). Without this, half-open WebSocket connections (caused by NAT timeout, firewall, or WSL network sleep) are never detected, and the SDK never triggers reconnection.

Configuration

channels:
  feishu:
    appId: cli_xxx
    appSecret: xxx
    pingTimeout: 30  # optional, default 30s

## Type of change

- Bug fix
- New feature
- Breaking change
- Documentation update
- Refactoring
- Other (describe below)

## Checklist

- pnpm run build passes
- pnpm run test passes
- New code has tests (if applicable)
- Updated documentation (if applicable)

[0xranx]

Thanks for the fix direction. I landed this in main with a small adjustment: the newer Lark SDK expects pingTimeout under wsConfig, not as a top-level WSClient option.\n\nAlso bumped the SDK peer range, added adapter tests, and documented channels.feishu.pingTimeout. Released in v0.48.1.