Move DSS to the end of each implicit stage

[dennisYatunin]

Purpose

This PR moves the DSS operation from the middle of each implicit stage (between the partial stage assembly and the start of Newton's method) to the end of each implicit stage. The change is applied consistently to all algorithms, and it ensures that stage values passed to our explicit and limited tendencies are always continuous across element boundaries.

The minor version number is also incremented so that we can immediately test this out in ClimaAtmos.

Motivation

For all simulations on a sphere, our metric terms (J and g) are discontinuous across element boundaries. Since the implicit tendency uses these metric terms, it introduces discontinuities to the Runge-Kutta stages. To avoid these discontinuities when evaluating explicit and limited tendencies, we must apply DSS after the last iteration of Newton's method (or after the linear solve in the case of Rosenbrock methods). We were previously applying DSS before Newton's method (and before the linear solve for Rosenbrock) because we incorrectly assumed that our metric terms were continuous. However, this intermediate application of DSS is unnecessary, since the implicit tendency does not involve any horizontal derivatives and can therefore handle discontinuous arguments. So, we can simply move DSS to the end of each stage in order to fix this issue.

---

• I have read and checked the items on the review checklist.

[charleskawczynski]

Did we test this in Atmos yet?

[dennisYatunin]

Opening a PR to test now.