Add Four lepton vertices to DAOD_PHYS

Hi everybody,

this MR includes several features required by the H->ZZ->4L analysis into DAOD_PHYS to add this ATLAS flagship as another user of the common format. The common features are:

1. The refit of the selected electron muon pairs to a common vertex. The vertices of all four-lepton combinations are saved to the FourLeptonVerices container. The covariances and the track weights are pruned from the containers to save a massive chunk of the needed disk space. The lepton selection criteria are documented in !56388 (merged). The physics performance of the vertex fit has been extensively validated

and the selection efficiency on the vertex quality at analysis level was found to be 99.95%. To assess the size increase due to the extra container, derivations from a data18 sample and several simulation samples were made. The table below summarizes the outcome from the checkxAOD macros. Even in the signal sample, the container needs roughly 0.01kB per event corresponding to a total size fraction of way below 1%.

Sample	#events	Mem Size	Disk Size	Size/Evt	Compression	Fracion of DAOD
Data	28862	5483.877 kb	149.372 kb	0.005 kb	36.713	0.%
ttbar	30000	5730.793 kb	166.952 kb	0.006 kb	34.326	<0.007%
qq->ZZ	30000	6215.175 kb	276.931 kb	0.009 kb	22.443	<1%
Higgs	40000	10776.627 kb	804.082 kb	0.020 kb	13.402	<1%

2. HTXS characterization model: This part adds several decorations to the xAOD::EventInfo in simulated samples to assess the Higgs production mode. The characterization is used by a wider range of Higgs analyses than the 4L channel.

• All samples: Add HTXS_prodMode decoration stating whether this is a signal sample of interest.
• List of samples mentioned in HiggsMCsamples.cfg: Full characterization of the Higgs events encoded in 18 integer, 8 float and 16 float vector decorations.

3. Information needed by the isolation close by correction tool: The idea of this tool has been floating around since the early days of the LHC. If two leptons from a Higgs decay are collimated enough, their associate tracks and calorimeter clusters would pollute the isolation variable of the other one. (Cf. the left plot below). The tool does nothing else than finding these pairs and subtract the energy of the associated id-track or Calo cluster (Cf. the right plot - please note that these one have been made ages ago, but they should satisfy the illustration purposes here)

I've opened another MR (!55961 (merged)) to update the tool to the area of release 22. It includes 2 algorithms which shall run at derivation level and do the following: * The IsoCloseByCorrSkimmingAlg searches for close-by lepton pairs. It adds the ID track of the lepton to the derivation if it pollutes the isolation variables of another lepton. The screen prompt below states the number of accepted tracks in the Higgs signal,

--- Muons: 2664 out of 70729 (3.76649%).
--- Electrons: 3261 out of 61003 (5.34564%).

the ttbar sample

--- Muons: 1275 out of 24658 (5.17074%).
--- Electrons: 833 out of 16634 (5.00782%).

data,

--- Muons: 1713 out of 15586 (10.9906%).
--- Electrons: 462 out of 8407 (5.49542%).

To correct the calorimeter isolation, the current plan is to decorate the electrons with the Eta/Phi/E of the associated Pflow elements. To ensure that the tool is actually run, another char decorator is added. For the muon case, we will likely add these decorations too, plus 2 extra ones containing the average eta/phi of the muon in the calorimeter. Maybe, we can get rid of the MuonCaloClusters then, if they are not already trashed in PHYS.

Tagging @goblirsc, @sabidi, @mdelmast, @tatsuya