It may be fragile, but it does sort of work.

This allows monitor() to be used for polling. Of course, you
probably want to use poll() first anyway, but it's nice not
to have to block if there's no output.

This is not as generic as having a single 'term' command, but
it's a lot more convenient to use.

This should make it easier for people who use screen for other purposes
to identify which sessions were created by mininet.

We may still need additional code to clean up after ovs.

There might still be some issues with this however...

For a network, we create "links" rather than "edges" on a graph.

"Edges" is a bit confusing because the links may not be edge links on
the network!

Also, since you're not necessarily running a "test", I changed it to
just say that we're "Done."

Now you can type commands into each mini-console!

The console tool spawns and monitors a bunch of nodes.

Now you don't need to specify a default read count.

Now, by default, we print the pid out in sendCmd() which allows us
to snarf it later and set lastPid correctly.

Now, we wrap os.kill() in a try block, so that if the process is
already gone we don't get an exception.