Respecting asymmetries in injections

[wbenoit26]

Re-filing the issue from the old repo. We should sample waveform kernels such that the coalescence point is allowed to be all the way at the right edge, but not at the left edge.

[EthanMarx]

To outline how this might look:

Instead of saving the waveforms centered, I think we add some right_padding argument that specifies how far from the right edge of the timeseries the coalescence lies. This will be saved with the signal dataset.

Then, the trigger_offset specifies how far from the left edge of the kernel the coalescence is forced to lie. We will force no constraint on the right edge.

When sampling kernels we'll need to pad the right edge with kernel_length - right_padding - trigger_offset worth of zeros. Then, we can sample start_idx such that the start_idx isn't too close to the coalescence time

[wbenoit26]

Yeah, I generally agree with that outline. We should include the ability to specify a distance from right edge for the case of signals where the defining signal time isn't at the end (thinking ahead a bit to the ringdown search).

I'm not sure there's a ton of benefit in not centering the signals, but I also don't have reason to not not center the signals, so I'm down to try out that parameterization. Regardless, the coalescence point of the signals is something that we ought to be saving as an attribute of the signal file.

[wbenoit26]

Closed by #158