Make nodes commit errors when the retrieval script panics

[tmpolaczyk]

When tally execution panics, the data request is resolved as "RadError::Unknown". I would expect a similar behavior if the retrieval or aggregation scripts panic, but that's not the case. Instead of commiting a "RadError::Unknown", nodes do not commit anything. So most probably the data request will resolve with an "insufficient commits" error.

This is similar to #2306, but this behavior has always been like that, ever since the creation of the RadManager. This is the piece of code that decides to ignore this case:

witnet-rust/node/src/actors/chain_manager/mining.rs

Lines 570 to 574 in 2fa9a2f

	.map_err(move |e| {
	// If resolving a data request results in a panic in the
	// message handler, ignore this data request
	log::error!("Couldn't resolve rad request {}: {}", dr_pointer, e)
	})

So not sure if this is intended or not. But we could even have nodes reporting the panic message, perhaps as a new RadError.

[aesedepece]

The behavior is different for tallies vs. commits because data requests need to be eventually tallied, so there shouldn't be such a thing as refraining from tallying.