Respecting asymmetries in injections

[alecgunny]

This will probably involve integrations with ml4gw, but I'm realizing more and more that our default behavior of centering injections and treating the trigger_offset symmetrically probably doesn't make much sense. We've talked before about how centering the raw waveform attaches the left tail of the waveform to the end of the coalescence, which is potentially problematic especially for lower mass injections. I've also been thinking that it makes sense to allow the coalescence to be closer to the right of the kernel than to the left, and that trigger_offset should maybe just refer to the latter.

I'm sure there are other places where this is crops up, but just generally it might be worth encoding explicitly the physical fact that the coalescence is not some center point of a symmetric signal but rather its end. Not sure how this might apply to other waveform types (e.g. I can imagine this is maybe not true for sine gaussians).