Should the histogram code automatically correct the PMF for unequal bins?

[mrshirts]

If there are nonequal width bins, then larger bins have more probability and thus a lower free energy. Example; if there are 100 samples, and 50 of them are uniformly between 0 and 90, and 50 are between 90 and 100, and there are two bins, they will have equal free energy, but the PMF should be lower in the 90-100, bin, since it has 10x the density of points.

If one subtracts -ln(bin_width) from each bin, the resulting free energies are a better match.

Right now, we expect people to do it themselves. But perhaps we should do it automatically.

[jaimergp]

Maybe an optional keyword? Something like correct_unequal_bins=True?