Fix inconsistent densities as a result of repeated changes of electron fractions in prob3

[thehrh]

Bugs were present in both Layers.weight_density_to_YeFrac() and Layers.weight_density_to_YnFrac(), where the attribute self.rhos_unweighted—never self.rhos—should have been rescaled. A test function demonstrates the reproducibility of the densities of the Layers object.

The code snippet in #868 now yields consistent event counts:

modell = DistributionMaker("IceCube_3y_neutrinos.cfg")
print(modell.params.YeM.value) # 0.4957 dimensionless
verteilung = modell.get_outputs(return_sum=True)[0]
count = verteilung.num_entries
print(count) # 39180.40671817049
modell.params.YeM.value = 0.5 * ureg.dimensionless
print(modell.params.YeM.value) # 0.5 dimensionless
count = modell.get_outputs(return_sum=True)[0].num_entries
print(count) # 39179.68692132214
modell.params.reset_all()
print(modell.params.YeM.value) # 0.4957 dimensionless (initial value)
count = modell.get_outputs(return_sum=True)[0].num_entries
print(count) # 39180.40671817049 (initial value)
count = modell.get_outputs(return_sum=True)[0].num_entries
print(count) # 39180.40671817049 (initial value)

Fixes #868