fix: add timeout handling for Daily room join/leave operations

[eroub]

Problem

The Daily transport's _join() and _leave() methods can hang indefinitely when the Daily completion callbacks fail to fire. This occurs because both methods use await future without timeout protection, waiting for the completion_callback to set the future's result.

When this happens:

• The leave() operation never completes
• The on_left event never fires
• Pipeline cleanup handlers are blocked

This issue was discovered in production where a call termination hung indefinitely at transport.leave(), preventing cleanup.

Solution

Add timeout protection to both _join() and _leave() operations:

1. Added module-level timeout constants (30 seconds each)

   • DAILY_JOIN_TIMEOUT_SECS = 30.0
   • DAILY_LEAVE_TIMEOUT_SECS = 30.0

2. Wrapped async operations with asyncio.wait_for()

   • Both methods now timeout after 30 seconds if the completion callback doesn't fire
   • Timeout errors are logged with room URL context for debugging
   • Error strings are returned following Daily transport's error conventions

3. Symmetric error handling

   • _join() returns (None, error_msg) tuple on timeout (matching its error signature)
   • _leave() returns error_msg string on timeout (matching its error signature)