Vertex-based luminosity lines

We need dedicated trigger lines that trigger on beam-beam and beam-gas vertices to cover the following use cases:

1. beam-beam and beam-gas imaging (for VDM and BGI calibrations)
2. ghost charge measurement (vertices in empty-empty crossings)
3. beam condition monitoring during physics data taking

To create such lines we need at least the following:

1. Primary vertex reconstruction (possibly dedicated) that is efficient for beam-gas vertices
   • covers a large Z range (+-1500 mm) with a worse efficiency expected at the extremes (2x-4x reduction in Run 2).
   • has a uniform efficiency as function as radial distance to the nominal beam line (beams pass at an angle, but also move during scans)
2. A PV-based selection algorithm that can decide based on the position and the track multiplicity of the reconstructed vertices. A selection based on forward/backward track multiplicity might be needed.
3. Combine the above into multiple lines
   • EE crossings for ghost charge
   • BE and EB crossings for beam-gas imaging
   • BB crossings for beam-gas imaging (excluding luminous region)
   • BB crossings for beam-beam imaging (potentially higher track multiplicity cut and/or prescale)

In addition, in Run 2 we ran the following as part of the vertex lines:

1. For every selected PV we ran the PVSplit algorithm that splits its tracks (semi-)randomly and refits two new PVs (informally called "split vertices").
2. The PVs and the "split vertices" were persisted within the HltSelReports as RecVertex objects, together with all their tracks. (Note that the more basic persistence of vertices in the HltVertexReports banks was not sufficient.)

This allowed to directly use the HLT1 reconstruction output for the BGI / VDM analyses, as well as for the online beam-gas monitoring task. This "Turbo for luminosity" approach proved to be very practical. However, if this is not available in Allen, we can resort to a CPU-based re-reconstruction in monitoring tasks / HLT2 / Offline.