Reduce the bandwidth and rates of RD lines

We currently have 634 lines in the RD TURBO stream. Assuming 0.5 GB/s of bandwidth dedicated to RD, as a rule of thumb we should be targeting a bandwidth smaller than O(0.001 GB/s)/O(1 MB/s) for each line (with a rate O(10 Hz)). This would bring us to a more comfortable O(0.5 GB/s) of output. This is a number that would fit into the LHCb constraints. The proponents of the trigger lines are encouraged to:

• Avoid persisting additional objects if they are not really necessary
• Reduce the rate of the lines by applying tighter requirements, if the persistency of additional objects in the event is crucial for the analysis
• Change the definition of how the additional parts of the event are persisted (e.g. the way to store candidates to compute the isolation variables)

Lines including \tau leptons seem to be the most problematic, more likely due to the big overlap they present among them and the presence of a large number of multiple candidates.

A possible alternative to select candidates for the isolation is the IPCHI2, which can be computed using the functor

F.IPCHI2.bind(F.FORWARDARG0, F.FORWARDARG1) < ipchi2_max