Increased Robustness: Prevent crashes if dataset contains only identical elements

[stheid]

In this case, (for example a degenerated probability distribution of 0 for all datapoints) the "abm" variant crashes during predict, because the logistic regression is called with the wrong amount of parameters

python/betacal/beta_calibration.py

Lines 115 to 120 in 3a9dbb4

	if self.map_[0] == 0:
	x = x[:, 1].reshape(-1, 1)
	elif self.map_[1] == 0:
	x = x[:, 0].reshape(-1, 1)
	return self.lr_.predict_proba(x)[:, 1]

This happens because the logistic regressions coefficients get 0 for such a degenerated case, and the lines above are executed accidentally.
De-aliasing the accidental 0 case with a None variable solves that issue.