Refactor extract starting topo

[mdtanker]

In run_inversion, prior to iterations starting, the function extract_prism_data takes in the supplied starting prisms_layer extracts the prism spacing, calculated the starting topography, and adds it as variable starting_topo and topo.

The way extract_prism_data calculates the starting topography is:

topo_grid = xr.where(prisms_ds.density > 0, prisms_ds.top, prisms_ds.bottom)

This seems like an unnecessary step which could have unintended issues. We should just make it a requirement for the user to add a starting_topo variable to the prisms_layer when it's supplied to the inversion.