Inncorrect use of simple method when random frames are selected in run()

[gitsirsha]

Thanks for the updates. I only have minor cosmetic changes.

While reviewing, I also realized that your code can solve a problem that has been inherent in the existing code: The run() method can take the frames=[0, 20, 25, 100, 1001, ...]` kwarg to select any frames instead of start/stop/step and if that's the case, the "linear" approach is incorrect (but that case is not being caught). We should select your new code path for un-even frame separations.

I am not asking for this to be resolved in this PR. If you agree that this is a problem and that your code could solve it, could you raise an issue for it so that it's not forgotten.

Originally posted by @orbeckst in #5066 (review)

This issue recognizes that while calculating msd of randomly picked frames (specified by kwarg by user in run), the newly added _conclude_non_linear() should be used instead of the previous _conclude_simple() method which is inaccurate when frames are non linearly placed. In most cases picking frames randomly results in picking frames that are non-linearly spaced so the _conclude_simple() method is not the right way to calculate msd.

[siddhardha2007]

Thanks for bringing this up! I agree that when passing frames=[...], using _conclude_simple() can produce inaccurate MSD results because it assumes linear spacing. In such cases, _conclude_non_linear() would indeed be more appropriate.