More meaningful control tests

I was testing some partial implementations of the control scheme. As feedback, I was using the signals returned by device or dynamic, but it turns out some of them are wrong. As a result the tests, though nominally working, are not really meaningful. These are the changes we identified as needed:

• dynamic.position should read device.robotState to (instead of default device.state)
• dynamic.velocity should (probably) read device.velocity (instead of default autoplug to 0)
• ZMP should be properly computed (separate issue, see #39 (closed))