Turned off PMTs with data-driven LCE

[terliuk]

When we produce simulations, we throw PMT hits according to the LCE of individual PMTs and then "zero" turned off PMTs with gain value. This apporach is totally valid as long as we have simulation-driven maps. But the moment when we turn to data-driven maps, the S1 LCE already includes not working PMTs by definition, but pattern maps do not. This can cause area/AFT/etc biases, so we have to think how to treat it.

Corresponding PR: #230

P.S. This issue was first accidentally put in private repo