Consider extending panic recovery to all Wasm critical errors

[guybedford]

Is there an existing issue for this?

• I have searched the existing issues

Description

The new panic recovery system recovers only on explicit Rust pancis.

There are other errors that would not trigger this recovery:

• Stack size exceeded
• Memory access out of bounds
• could not allocate memory
• unreachable

It could be possible to also catch these cases and similarly issue a Wasm VM reset internally, by wrapping all top-level invocations with these error handlers.

This would be a bit more invasive than the panic handler work though, so should only be investigated if we are seeing other errors of this type transitioning Wasm into an invalid state.

[guybedford]

Supporting out of memory would be the highest priority here to start at least.