Maybe `time_dep_op(t)` should not mutate

[amilsted]

For time-dependent operators we have set_time!(op, t) to set clocks in place (and return the updated operator).

Currently op(t) is equivalent to set_time!(op, t), which could easily be a bit unexpected. For example op(0.0) == op(1.0) will return true always, since the LHS and RHS are refer to the same struct.

Instead, consider having op(t) return a copy (and make sure we are using set_time! in time evolution code.