[ENH] Improve coverage

[TonyBagnall]

Describe the feature or idea you want to propose

One of my next priorities will be to get coverage over 90% on the overnight run. It is on 83.81 on 16/8/23, with 40324 out of 48114 lines covered. Thats nearly 8000 lines of uncovered code. I'm really hoping there are some easy wins here!

https://app.codecov.io/gh/aeon-toolkit/aeon/tree/main/aeon

By module, the most missing lines (and that module's coverage) transformations: 1202 (86.97) forecasting: 810 (87.86) classification: 779 (82.97) datasets: 586 (55.90) utils: 430 (85.44) datatypes: 388 (82.07) benchmarking: 326 (48.89) regression: 225 (80) base: 216 (77.47) annotation: 161 (85.44) distances: 137 (88.78) networks 134 (66) clustering: 99 (88.05)

Note none are over 90 atm. For context, scikit learn has 97%. I will probably first look at datasets and benchmarking

Describe your proposed solution

write tests to cover uncovered code, esp testing exceptions etc. Look for easy wins first

Describe alternatives you've considered, if relevant

No response

Additional context

No response