Simplify the "is a character ready" test, so that we don't return a false
negative for keypresses with zero ascii values. This is in line with the
comconsole test, rather than being more ambitious.

Submitted by: rnordier