light yield fix for Run-2

[mvicenzi]

Note : This is being opened against feature/pmtsim_gain to showcase meaningful differences for review, but it should be rebased on the production branch once #892 is merged.

This PR introduces two empirical/effective corrections that close the light model discrepancy for Run-2 by fixing both the signal shape and the overall PE normalization (see SBN-docdb-48157, SBN-docdb-48511).

Two corrections are introduced:

• ApplyTailSuppression() applies a subtractive, causal correction to the final waveform. It tracks a running low-pass-filtered estimate Q of the waveform (filter with time constant Tau) — and subtracts Eps x Q from each sample (clamped so the pulse cannot flip polarity). Esp is a factor that controls the strenght of the correction. This removes the excess integrated charge in the tail while leaving the prompt peak essentially untouched.

• Distance-dependent photon survival probability: instead of a flat QE cut, each scintillation photon is now kept with probability QE x S(d), where S(d) is a function of the distance between the emission point of the photon and the PMT. It is configured via specifying distance bins via DistanceSurvival.BinEdges and the corresponding values with DistanceSurvival.Factors. It defaults to 1 outside the configured bin range.

Both corrections are currenlty enabled and tuned ONLY for Run-2 data.
The tune for Run-3/4 is still being optimized.

[PetrilloAtWork]

I left some minor comments.
This is a configuration-breaking change, and I mentioned a way not to make it so.
However, this PR is very well written and very well documented.

[mvicenzi]

Thank you @PetrilloAtWork !
I have implemented your suggestions.