Add homogeneous source option.

Mostly useful for testing. Removes the field/potential related source terms from the evolved system.

This should probably mostly be used with nonad_zero = F and this will then remove the potentials from the collisionless linear time advance. This should lead to the response matrices being zero prior to inversion and hence the resulting potentials are not accurate when used with this option. As such, operations/diagnostics which rely on the potentials will also not be correct. This includes anything which calls g_adjust such as the collisions and some of the moment diagnostics. This could be rectified by calculating the potentials directly from the distribution function at the end of the time step using a Maxwell solve (e.g. calling dist_fn::get_init_field).