Fix nyquist plotting from FrequencyResponseData

[wueestry]

This PR resolves the issue #992, by adding a check for FrequencyResponseData during plotting of a nyquist plot and setting omega if not yet set correctly

[coveralls]

coverage: 94.497% (-0.007%) from 94.504%
when pulling 7fa5960 on wueestry:nyquist_frd
into e1d21d2 on python-control:main.

[murrayrm]

This is a good fix, but I think we could add a bit better support if there are multiple FRD systems with different frequencies and not explicit frequency list given. For example, try this:

import control as ct
sys = ct.rss(4, 1, 1)
sys1 = ct.frequency_response(sys, np.logspace(-1, 1, 10))
sys2 = ct.frequency_response(sys, np.logspace(-2, 2, 10))
ct.bode_plot([sys1, sys2])                     # works OK
ct.nyquist_plot(sys1, np.logspace(-1, 1, 10))  # works OK
ct.nyquist_plot(sys2, np.logspace(-1, 1, 10))  # generates error (as it should)
ct.nyquist_plot([sys1, sys2])                  # generates error (does it need to?)

It seems to me that the last case should work OK and just plot each system independently, without needing a common frequency range (as it does for Bode plots).