Question the need for H_np and U_np

[dngoldberg]

In solver/timestep(), H_np and U_np are assigned to H and U after they are updated (this is an operation which must be annotated). U_np and H_np are used to define the RHS in def_thickadv_eq(). They do not seem to serve any other purpose (other than that sometimes diagnostic output uses U_np and H_np). Looking over model.py does not seem to contradict this (though possibly they are necessary for operations in the runs/ folder).

Are they necessary, and if not, is it an issue to keep them (or could it lead to potential confusion)?

If they are necessary for forming the RHS in def_thickadv_eq(), there is an inconsistency in line 515:

fl_ex = self.float_conditional(H)

since fl_ex is used in the RHS definition (at line 549).

[jrmaddison]

It seems these can be removed -- see #61