Count flops proportional to number of stages etc in RK implementation

For vectors of non-trivial size these additional counts are completely swamped but in PETSc we have traditionally tried to track such flops so I felt obligated to include what I could.

Commit-type: housekeeping