[ENH, BUG] Distance refactor and bug fix

[TonyBagnall]

A range of internal changes to distances module and how they are used. The primary change is in the internal representation. From the beginning, we assumed an instance was of shape (m,d), where m is the series length and d is the number of dimensions. The reason for this was because thats how others did it, and there were some native python operations that were more efficient. However, it was later decided to use and (n,d,m) throughout sktime for numpy representations (n is number of cases) . This meant any estimator using the distances would need to transform internally, since it is not possible to know if data is passed in (d,m) or (m,d) shape. This has caused errors in code (e.g. KNN was broken for a while in 2021 because the transform was removed) and its generally far too brittle. This PR converts all distances to expect input series to be shape (d,m). Given it is all compiled to C, there will be no performance hit, its all just loops in the end. This means we can remove all transposing from distance clusterers and classifiers.

This PR has got a bit too big, apologies. A few other things includes

1. Classifiers and clusterers using difference series now have option to taking the difference just once. direct calls to ddtw and wddtw still take the difference of the series, but kmeans takes the difference once and switches from ddtw/wddtw to dtw/wdtw
   This has been done by introducing another function average_of_slope_transform in ddtw rather than as a transformer. We could make it a transformer if thought desirable
2. Add doc strings to describe distances. Let me know if any of these are unclear. I could have added more mathematical description for some. However, I have written the maths in sphinx, but I currently have no way of locally debugging it, so not sure what it looks like. I've put references for each.
3. Add correctness tests for distances. These should be restructured to conform to standard parameterised approach, but would rather raise an issue and do that after the thing is in, would be a good first issue
4. Single problem loads unit_test and basic_motions now have option to return numpy, but no other changes in the problem loaders, so hope this is not controversial
5. fixes a bug in EDR and ERP, both of which were not using the correct data in calculations due to a typo in a nested local function
6. fixed a bug in numba conversion that was still going to (m,d) on edge cases
7. fixed pairwise_distance to remove transpose
8. remove all transposing from distance based classifiers and clusterers

[TonyBagnall]

note to self, wait until after Franz does pydata thing to merge this

[fkiraly]

note to self, wait until after Franz does pydata thing to merge this

Indeed, the pydata thing is done now 😄