Added steps=n logic for skipping steps

[alessandrofasse]

See https://github.com/patrick-kidger/diffrax/issues/

[patrick-kidger]

Thank you for the contribution! I like the look of this. I've left a first round of review comments. :)

[patrick-kidger]

We should also add tests for SaveAt(steps=n, t1=True) and SaveAt(steps=n, t1=False) when the number of steps does not divide n. Then check that in the first case with e.g. n=2 and 5 steps that we store [0 2 4 5] and in the second case we store [0 2 4]

[alessandrofasse]

Dear @patrick-kidger ,

I continued on the pull request and I could fix most of the cases. However, I am now stuck in an issue were are a couple of hours I could not find the problem.

Im shorts its about _save_t1 function. I am not able to get this right and now several tests were failing. I added the following tests, as you suggested, and I think my assert would be correct (can you cross check?)

Can you maybe assist me here a bit, I think I am misunderstanding about the logic on how states are stored and logged anlong the way.

You help would be much appreciated! 🙏