Optimise charm neutral lines

Charm neutral lines (three-body with pi0 and radiative decays) take up a significant fraction of the charm bandwidth:

• d0_to_hhgamma.py -> ~250 MB/s
• d0_to_hhpi0.py -> ~180 MB/s
• d_to_hhhgamma.py -> ~35 MB/s

but are still missing the information to apply an improved ECAL calibration or gamma/pi0 reconstruction offline, which could be crucial to optimise the S/B ratio etc. This MR tries to reduce the rate of the lines wherever possible, and at the same time it adds the information on the ECAL deposits for the signal and surrounding ECAL region. Therefore, the rates will decrease, but the event size will increase.

Main modifications are:

• adding a DOCA < 100(150) um requirement for decays into two (three) tracks. This is checked to reduce the rate by 10% on D0 -> hhgamma
• persisting the ECAL information for neutral candidates (possibly including isolation particles). This increases the bandwidth by around 30% in D0 -> hhgamma
• selecting only K- pi+ pairs within 120 MeV from the K*(892) mass for the D0 -> K- pi+ gamma line (reduces the rate by around 50%)

Possible changes in case the rate needs to be reduced further (left for a future optimisation):

• Regarding selective persistence, one could save only tracks compatible with coming from the D decay vertex if the main goal is only rejecting partially reconstructed D decays. The value of the cone aperture for the neutral particles should be also rechecked and tuned to the analysis requirements
• Add a prescale for D0 -> K- pi+ pi0 (its BR is larger than that of D0 -> pi+ pi- pi0 by around a factor of 10). However, there are ongoing studies to use this decay as a control channel for measuring ACP(KK), due to a better kinematic overlap with the D(s)+. Therefore, this change is not applied yet.

FYI @jbrodzic @smaccoli @ldufour @fbetti