truncate calo digit energy to multiple of 0.2 MeV when computing position

In order to make the Calo reconstruction more reproducible (eg. less dependent on digit order) truncate the energy of individual digits to 0.2 MeV.

To set the scale (thanks to @jmarchan and @yamhis):

The Ecal is mostly calibrated in transverse energy, and the current 12-bit configuration provides uniform Et dynamics up to a saturation point of 10+7sin(theta) GeV (the PMT gain varies with the distance to beam): 1 ADC = 2.5 + 1.75 sin(theta) MeV (2.5 - 3 MeV)

And the electronic noise is ~1.2 (incoherent) + 0.3 (coherent) ADC. A 3x3 cluster has a noise of ~6.3 ADC=15-20 MeV

so this truncation is (at least) an order of magnitude below the ADC precision.

Requires Brunel!990 (merged) and Moore!449 (merged) (ref updates)