Arrange for oneCmd.py to return keys from subcommands and broadcasts.

oneCmd.py only spits out keywords returned by the command sent to the top-level actor it calls. It should also return keywords from subcommands, and optionally from broadcast commands during the time the top-level command is active.

Example: the output from mcs expose bias does not include the interesting output from the gen2 actor.

The output is purely informative: this is just to make life easier.