move optmap application to processor

[ManuelHu]

• in hit processing step
• split the lar (scintillator output scheme) table to individual SiPM channel tables
• either by using the current multinomial approach, or by finding a nicer way to do the statistics (i.e. only loading one channel's map at a time)
• output hit tables S0xx with evtid, times
• future: i.e. SiPM model as hit processor after the splitting processor

---

• keep alternative to just process the LAr table (i.e. to cut on edep, and not on p.e. hits)

[tdixon97]

Regarding the multinomial.
Instead of the distribution being:

$$ p(\vec{n}|\vec{\mu}) = \text{Pois(}\sum n,\sum \mu) \times \text{Mult(}\vec{n}| \vec{p}, \sum n) =\prod \text{Pois(}n_i|\mu_i)$$

This is the same as how a binned likelihood can be either written as a product of a multinomial and an overall poisson, or a product of possion. I think this also applies in this case
Probably you can find the derivation in many stats books, but I can't be bothered.