[MNT] Testing of kshape after `tslearn 0.7.0`

[hadifawaz1999]

Describe the issue

After the release 0.7.0 of tslearn, the small change in kshape here is making our test fail because the "Expected results" should be updated on basic motions dataset here

Suggest a potential alternative/fix

update expected results

Additional context

No response

[MatthewMiddlehurst]

Currently has a lower bound from #3107 for CI, but still needs to be resolved. Should not be that difficult.

[Nithurshen]

I also updated the ROCKAD docstring in aeon/anomaly_detection/series/distance_based/_rockad.py. The test was failing due to a minor floating-point precision mismatch (at the 8th decimal place, from 0.0699094 to 0.06990941) caused by the dependency updates. I updated the expected value to match the current output.