[ENH] Rationalise matrix profile transformers

[TonyBagnall]

Describe the feature or idea you want to propose

there are two transformers that implement the MatrixProfile. The one in transformations.collections is a bespoke implementation, the one in transformations.series uses (an incorrectly handled) soft dependency. There can be only one

There is also a function MPDist in distances, which uses bespoke operations that are native python. Could be optimised.

Describe your proposed solution

Implement everything with stompy.

we see if they are indeed the same, profile one and then assess alternatives for including MP as a transform.

Describe alternatives you've considered, if relevant

optimise the current implementations

[TonyBagnall]

Collection transformer: MatrixProfile
https://github.com/aeon-toolkit/aeon/blob/main/aeon/transformations/collection/_matrix_profile.py Bespoke implementation
Series transformer: MatrixProfileSeriesTransformer
https://github.com/aeon-toolkit/aeon/blob/main/aeon/transformations/series/_matrix_profile.py
Distance: mp_distance
https://github.com/aeon-toolkit/aeon/blob/main/aeon/distances/_mpdist.py
Similarity search
https://github.com/aeon-toolkit/aeon/blob/main/aeon/similarity_search/matrix_profiles/naive_matrix_profile.py
https://github.com/aeon-toolkit/aeon/blob/main/aeon/similarity_search/matrix_profiles/stomp.py