feat: WeCom Stream Mode (WebSocket long-connection)

[felix307253927]

Summary

Add WeCom Stream Mode (WebSocket long-connection)

Changes

Testing

• cargo clippy --workspace --all-targets -- -D warnings passes
• cargo test --workspace passes
• Live integration tested (if applicable)

Security

• No new unsafe code
• No secrets or API keys in diff
• User input validated at boundaries

[jaberjaber23]

LGTM. Legitimate WeCom Stream Mode channel adapter. Clean 822-line implementation with proper WebSocket protocol handling (subscribe, ping heartbeat, msg/event callbacks), message deduplication via ring-buffer cache, Zeroizing for secret, outbound message queue, and correct ChannelAdapter trait implementation. Config integration is clean. Please rebase on main.

[jaberjaber23]

This PR has merge conflicts. Please rebase onto the latest main branch and resolve conflicts so we can merge.

[jaberjaber23]

Thanks @felix307253927 — WeCom stream mode via the official WebSocket endpoint (wss://openws.work.weixin.qq.com) is a real feature request. Branch is currently CONFLICTING / DIRTY against main after #1041 landed.

Before re-review:

1. Rebase on current main and resolve the Cargo.lock / channel-module conflicts.
2. Verify token/secret handling across reconnects. The WebSocket client is long-lived — confirm no secrets are logged on reconnect attempts, auth failures, or frame decode errors. Zeroizing on the secret is good; please also audit any tracing::info! / error! calls in the reconnect loop for accidental inclusion.
3. Add tests for: initial auth failure (bad secret), mid-stream disconnect + reconnect, and backpressure when the channel buffer fills.
4. Security Audit was red on the previous run — rebase should help; must be green for merge.

Happy to re-review once rebased.

[jaberjaber23]

6 files including mcp.rs for what's titled WeCom Stream Mode. Why does the MCP runtime path change for a channel adapter? Please either remove that hunk or split it into its own PR with a clear explanation.

Also 40+ days old now and likely conflicts with main. Please rebase, drop the mcp.rs change unless it is genuinely required by the WeCom path (with rationale in the PR body), and we will re-review.

Extra scrutiny applies because of the cmd /C safe-path regression caught in your earlier PR #749. Clean focused submission is the path to merge.