fix(mcp): bound image uploads and expose bridge diagnostics

[lizhong1989]

Summary

• bound uploadImage and uploadImage:complete requests to a dedicated 60-second timeout
• expose Primary bridge PID, start time, and uptime through /status
• show Primary process diagnostics when a Secondary CLI instance times out
• reject pending requests when the bridge stops
• document WECHATSYNC_UPLOAD_TIMEOUT

Why

A platform adapter can leave uploadImage(blob) pending indefinitely. The CLI processes local images sequentially, so one stuck upload blocked the entire sync until the bridge-wide six-minute timeout.

Separately, when an old Primary process still owned ports 9527/9528 but the extension was disconnected, new CLI instances could only report a generic connection error. Exposing process diagnostics makes the stale owner identifiable without introducing unsafe automatic process termination.

Closes #208.
Partially addresses #207. A safe reset/takeover mechanism still needs an ownership or heartbeat design.

Tests

• corepack yarn workspace @wechatsync/mcp-server build
• corepack yarn workspace @wechatsync/cli build
• corepack yarn workspace @wechatsync/mcp-server test

Added coverage for:

• Primary status diagnostics
• bounded image upload timeout
• rejection of pending requests during bridge shutdown

All 3 tests pass on Windows 11 / Node.js 24.14.0.

Compatibility

• syncArticle keeps the existing six-minute timeout.
• Existing /status fields remain unchanged; new fields are additive.
• The default upload timeout can be overridden with WECHATSYNC_UPLOAD_TIMEOUT.