BUG: RLM fit with start_params raises if only one parameter

[josef-pkt]

if start_params = np.array([5]) one element, 1-dim array, then the squeeze removes the 1-dim which then raises exception in shape check

statsmodels_gh\statsmodels\statsmodels\robust\robust_linear_model.py in fit(self, maxiter, tol, scale_est, init, cov, update_scale, conv, start_params, start_scale)
    264         else:
    265             start_params = np.asarray(start_params, dtype=np.double).squeeze()
--> 266             if (start_params.shape[0] != self.exog.shape[1] or
    267                     start_params.ndim != 1):
    268                 raise ValueError('start_params must by a 1-d array with {} '

IndexError: tuple index out of range

I ran into this for the constant only regression for resistant estimators in #9227

I don't know why there is the squeeze.
If we want or need to keep the squeeze, then a fix is to add
start_params = np.atleast_1d(start_params),

I will fix it this way in my PR, where I need it now