fix: inverted axes for variances of 2D weighted histograms when transformed to hist

[ioanaif]

No description provided.

[jpivarski]

As I understand it, you're applying the same fix to _values_variances as you've applied to to_boost, and the PR will be done when that has been tested at least once locally.

[ioanaif]

In TH2, _values_variances does perform the axis swap, thus all is good in the computation of the variances.

I removed both the previous computation and my fix to the computation from to_boost and replaced it with simply calling .variances(flow=True) since variances are already computed at that stage.

This should fix any problem with the variances' computation and remove duplication for easier debug in the future.

It's ready to go imo. ☺️

[jpivarski]

Okay, great!

I saw that you're using self.variances, which runs self._values_variances, and so the values are being computed twice. (That's the original reason why this function reimplemented it.)

But we can arrange to have values computed only once by doing self._values_variances in the if-clause and self.values in the else-clause. In both cases, flow=True.

I think this is ready to go, but since I made an edit, I'll give you the chance to counter-edit before merging. Go ahead and merge when you think it's ready.