Fix for bug in margin.py about python-control

[cwrowley]

Good catch--thanks @callawaycass

Now that you pointed this out, it seems inconsistent for this routine to take phase inputs in degrees, while sys.freqresp(omega) returns phases in radians. For instance, one might expect the following code to work:

sys = TransferFunction(1,[1,1])
mag, phase, omega = sys.freqresp(np.logspace(-1,1,100))
margins = stability_margins((mag,phase,omega))

but the units of phase are wrong. So I wonder if we should change freqresp to return phase in degrees, in order to be more consistent? I notice the docstring for freqresp doesn't specify the units of phase either, so regardless of whether we change the routine, we should improve the documentation.