A possible edge race condition between an example that it is hung,
byexample sending an abort (ctrl-C) but in that moment the example
finishes and the signal is received by the interpreter that finishes
also leading to an InterpreterClosedUnexpectedly.

We cannot "fix" the RC but we can handle the case and print a message
instead of an exception.

Some interpreter may have a race condition between the output generated
by the example and the prompt printed by the interpreter.

In those cases the output may appear *after* the prompt.

This not only breaks the example (the last bits of the expected output are
missing) but it also breaks the next example (unexpected bits are at the
begin of the expected output).

Assuming that the timming issue is quite small, we can add an optional
delay after the prompt to capture more data at the end and try to keep
the interpreter and byexample in sync.

The interpreter for Rust, evcxr, works quite well but it has a high
runtime: 10-30 seconds for starting up the runner and 1-4 seconds for
running each example.

For now it is in experimental phase.