Double panics can still happen in the wasm node

[tomaka]

When a panic happens in the Wasm node, we call throw, and intentionally let the JavaScript worker die as a result.

This is totally intended. A panic could leave memory in a corrupted state, and calling functions again is undesirable.
See this paragraph for context.

It seems that, despite the worker errorring, the timeouts and sockets that are still alive in the worker continue to run and continue to call the smoldot Wasm again.

[tomaka]

To reproduce: put a panic!() somewhere in a background task, then observe how there are two panics instead of one.

The second panic (something regarding WouldBlock) comes from the fact that we call try_lock on a mutex that isn't expected to be locked but is still locked because of the first panic.
The proper fix is that we never reach again Wasm code again after the first panic happened.