Revise snow probability computation

[guidocioni]

At the moment snow probability is computed in the same way precipitation probability is computed here

point_wx/src/pages/ensemble/figures.py

Lines 210 to 217 in b3cf037

	df[f"{var}_prob"] = (
	((df.iloc[:, df.columns.str.match(columns_regex)] >= 0.1).sum(axis=1) / members)
	* 100.0
	).astype(int)
	df[f"{var}_mean"] = df.iloc[:, df.columns.str.match(columns_regex)].mean(axis=1)
	df.loc[df[f"{var}_prob"] < 5, [f"{var}_prob", f"{var}_mean"]] = np.nan

So we take the fraction of ensemble members that have snowfall >= 0.1cm.

I think we should instead interpret the snow probability as fraction of members that have precipitation AND this precipitation is frozen.
So we should only consider the members that have precipitation >=0.1mm, and over these members see which members have snowfall. This should give a more appropriate estimation of snow probability.

Also, the logic of computing both probabilities I think should go into the get_ensemble_data function instead than sitting in this function that is only used to create the figure.