Future version of smooth evolution of the coupling

[mabuni1998]

In the solver, the current way to set the time index of the operators is:

WaveguideQED.jl/src/solver.jl

Lines 29 to 33 in 2e4d63c

	function get_hamiltonian(time,psi)
	tidx = min(round(Int,time/dt,RoundUp)+1,nsteps)
	set_waveguidetimeindex!(ops,tidx)
	return H
	end

Since the solver often uses timesteps much smaller than the timestep defined in the basis, the input is often between two-time bins. This is handled by dividing the input (float) with timestep (another float) and rounding. Thus we infer which time bin the input time corresponds to. This is faster(?) than looking into the whole array of times and seeing which two-time bins the input time lies between. However, dividing two floating numbers and turning them into an index seems like bad practice. Any inputs on another implementation @Krastanov ? (the current one works, although I recently had to put min() in because floating point errors made the last index go beyond the maximum time index).