Prefer filters over selections

Optimise thread usage by using ODIN event list filters rather than ODIN selection, as discussed in #499 (closed).

No change in the selected number of events is expected and indeed, by running the hlt1_pp_lumi sequence on master and on this branch, the enclosed logs are found. In particular, we have for both

Hlt1ODINLumi:                                  4192000/8000000, (15720.00 +/-     5.30) kHz

Some throughput is saved (1.5% on MC, but a larger gain might be expected on data as MC doesn't test e.g. the orbit and nobias lines) closes #499 (closed)

Also, this adds a line selecting ee bunches far from activity (as tagged by ODIN) that are useful for noise estimation (as required in https://gitlab.cern.ch/lhcb/opg/-/work_items/55)

cc @cagapopo to ask a review :)
and lumi conveners @fferrari @elniel @rsadek and @dcraik as this is dealing with luminosity

closes #499 (closed) closes https://gitlab.cern.ch/lhcb/opg/-/work_items/55 goes with MooreOnline!337 (merged)

hlt1_pp_lumi_odin_branch.txt hlt1_pp_lumi_master_branch.txt