Added 'level_order' for custom ordering in stacked_barplot

[redst4r]

Hi,

just added an option to customize the order in which the different levels (x-axis) in stacked_barplot() are plotted.

Ideally, the ordering would be encoded using pd.Categorical for the data.obs[feature_name] column and its inherent order, but the plotting code is currently agnostic of that.

[johannesostner]

Hi! Thanks for this PR!

I really like this idea, allowing people to change the order of bars within the plotting function is often much more convenient than sorting the dataset beforehand.

I would also like to incorporate pd.Categorical's inherent ordering, as you mentioned. The ordering priority for a categorical covariate should be level_order > inherent order > order in the dataset.
For this, I will switch the target of this PR to our current release branch, and merge it there. The feature will then be released with the 0.1.2 version.