Uniformize heat/hot spell indices

[coxipi]

Generic Issue

• xclim version:
• Python version:
• Operating System:

Description

• Currently, hot_spell_max_length has window=1 by default, wherehas hot_spell_total_length and hot_spell_frequency have window=3. I suggest we change this to 3 everywhere, so that hot_spell_XYZ describe the same events, and just a different diagnostic
• heat_wave_index is just hot_spell_total_length with different defaults. "heat" is usually used when both tasmin,tasmax. We could remove this index since it's redundant? At the very least, I will use hot_spell_total_length for the computation
• heat_spell_XYZ indicators (in _temperature.py) basically do the same as heat_wave_XYZ, we could keep the latter.

Code of Conduct

• I agree to follow this project's Code of Conduct

[aulemahal]

heat_spell_XYZ do some arithmetic over the spell while heat_wave_XYZ do not (the criterion is that all days pass). The difference was kept for three reasons : 1) I needed a different name in PC, 2) I didn't want to break anything, 3) I was in a rush and didn't have time to think.

Thus, I think it's ok to remove the heat_wave_XYZ indices, but we could keep the indicators ?

[coxipi]

Ah right, rolling averages. So heat_wave_XYZ is the bivariate version hot_spell_XYZ , whereas heat_spell_XYZ does its own thing.

Well, heat_wave_XYZ indices could probably be removed by using bivariate_spell_length_statistics and win_reducer = "min" ? A lot of indices could be removed if we follow this logic

My main point was the first two points, I just added the comment about heat_wave after, sorry I read too fast. Similarly, we should not change anything about these points to avoid breaking changes? We can keep heat_wave_index as an artefact I guess, but the fact that heat_spell_XYZ defaults are not coherent is a bit more bothering I find.

[aulemahal]

I think so indeed. It is the goal of having these feature-full "generic" indices that we can remove duplication and jump straight from the "generic" to the indicators.