Implement waveform simulation and hit finding algorithm for mPMT

[kmtsui]

This PR should be merged after PR #7

---

This PR implements the hit finding algorithm as described in issue #8
In addition, the simulated waveform and the Q and T pulls (digi minus truth) can be saved in the output file.

Waveform simulation

Implementation in HitDigitizer_mPMT::BuildWavetrain()
For each photon hit, the expected voltage is sampled at a 8 ns interval.

There is no variation in the 1PE pulse shape, but the sampling time (relative to t=0 in the 1PE pulse shape) is dependent on the photon detection time, which introduces randomness in the sampled waveform. Below are the few examples.

If there are multiple photons arriving a PMT in a short time, the waveforms are added together. Below are the examples of a 2-photon (left) and 4-photon (right) waveform.

Hit finding algorithm

Implementation in HitDigitizer_mPMT::FitWavetrain()
Each waveform is scanned to find the local maxima that satisfy trigger conditions. Then digi time is estimated using Constant Fraction Discriminator, and digi Q from the sum of eight samples around the maximum.
Below are the pull distributions (digi minus truth) from a 156 MeV kinetic energy pi+ beam particle simulation in WCTE. For a multi-photon hit, true time is defined as the first photon detection time, true charge is the number of photons, in a time interval of (-1, +8)*8ns around the digi time.

Time pull is usually small. Meanwhile the digi charge can be off by more than 1 unit, due to photons arriving at a time larger than the charge integration window but less than the dead time period. Figures below show such waveforms.