KernelSHAP raises a warning if output dim is 1 and task is regression

[jklaise]

This could be confusing for the user as simple regression is a valid use case:

alibi/alibi/explainers/shap_wrappers.py

Lines 692 to 696 in 7d73473

	if not self._explainer.vector_out:
	logger.warning(
	"Predictor returned a scalar value. Ensure the output represents a probability or decision score "
	"as opposed to a classification label!"
	)

Same applies to the open TreeSHAP PR (#206).