Better dpph timebase

[laroque]

Currently the dpph script makes some hard-coded assumptions about the resonance width and also lock-in sampling.

It would be better if:

• [x] The number of points were configurable (it is)
• [x] The sampling interval were determined from the sweeper and the requested number of points (currently it isn't, this involves asking the sweeper for its number of points and dwell time, then adjusting the sampling time based on the lock-in's number of points such that the lock-in data taking over-covers one sweep... I like 110%)
• [x] The width of the target should be stored not as a number of samples, but as a frequency which takes into account both the lock-in's sampling rate and the sweeper's sweep rate.
• [x] The target width should be an argument with a (smart) default value, not something hard-coded into the script
• [x] Also, the target is currently the derivative of a gaussian (because that is easy). It should actually be the derivative of a lorentzian. This shouldn't make much difference but would be more correct so we should do it.

[laroque]

Note that the lorentzian filter seems to have a much stronger response (amplitude of the overlap is 2-4 times as large as a gaussian target function). The large tails do result in the filter response being more than zero over a decent fraction of the frequency range, but the highest features appear to increase by a significantly smaller factor than the signal itself, so I expect that these results are more robust.

I've merged these changes into develop. The dpph script could still use quite a bit of cleanup, but since it is working I'm going to leave that very low on the todo-list.