Potential improvements to inner loop of SIQN timestepper

[tommbendall]

There are some changes we could explore with the inner loop of the SIQN timestepper:

• updating the reference profile, as suggested in we currently don't update the reference state in SIQN (unlike LFRic) #484
• setting the density and temperature residuals to zero when not in the first inner iteration (as is done in GungHo)
• explore setting xnp1 to be xp+xrhs_phys for the first outer iteration and inner iteration

[tommbendall]

This will have been addressed by #536, #539 and #542