calculate_residuals wrong point

[ChrisRackauckas]

In some of the calculate residuals, it's using u1-u0 and then using the max of those two in the denominator. However, it should be using the difference for an error estimate, and then the max of uprev and u together. Since this is violated, the stability of the non-stepped method then matters for the stepping. Not a huge deal but worth documenting.

@devmotion do you remember some discussion about this?

[devmotion]

Hmm no I don't remember any discussion about this.