improve forced_response and its documentation

[bnavigator]

• Fix doc error for return_x as revealed by control.forced_response argument T #586.
• Remove redundant kwarg defaults. They are in the call signuature. Numpydoc compliant default indication
• Enhance error message if ctime system without timevector is provided
• Make sure the returned time vector keeps it's length

[bnavigator]

Strange that the markov test fails here too.

https://github.com/python-control/python-control/pull/588/checks?check_run_id=2188093854#step:5:212

A rerun succeeded. Wonder what triggered the length mismatch of Y and U.

[coveralls]

Coverage increased (+0.2%) to 89.448% when pulling a8b72f5 on bnavigator:timeresp-doc into f1a9860 on python-control:master.

[bnavigator]

Even without the modifications of this PR, testMarkovResults started to fail sporadically (#588 (comment)), because scipy.signal.dlsim returns not enough samples when the last element of the time series is just below a multiple of sys_dt

Fixed with a6ea296

TODO: cover the new lines. They were only triggered by the testMarkovResults test depending on the random seed.

[sawyerbfuller]

Looks good. Seems reasonable to allow a nonzero start time!