[MCP browser-automation] Improve timeout error messages with diagnostics

Problem

When browser_wait_for_selector times out, the error message is too generic to debug:

MCP error -32000: Tool failed: Request timeout

This doesn't indicate:

• Whether the element was ever found
• The current page URL
• If element exists but wrong state (hidden vs visible)
• The selector that failed

Suggested Improvement

Include diagnostic information in timeout errors:

MCP error -32000: browser_wait_for_selector timed out after 30000ms
- Selector: .source-item:not(.loading)
- Page URL: https://notebooklm.google.com/...
- Reason: Element not found on page
- Expected state: visible

Helpful Diagnostics

1. Selector - What we were looking for
2. Page URL - Where we were looking
3. Failure reason - Not found / found but wrong state / MCP connection issue
4. Element state - If partially matched (e.g., "element exists but hidden")
5. Elapsed time - How long we waited

Impact

Better error messages would significantly improve debugging efficiency for browser automation tasks.

Related

• https://github.com/trungnguyenvn/claude-code-kit/issues/1