Data model fix: avoid classifying TPC loopers as physical primary

[ddobrigk]

This PR re-defines mcparticle::IsPhysicalPrimary to embed a safety feature in which particles that are (a) created beyond 5 cm from the beam axis and (b) marked as physical primary will be forcibly re-classified as FromBackgroundEvent. The new definition is provided in a separate namespace called mcparticle_v2, given that mcparticle::IsPhysicalPrimary is used multiple times in derived data.

Further, this PR adds a layer of abstraction to mcparticle::Flags by declaring two new columns: one standard mcparticle_v2::Flags (getter: storedFlags()) and a dynamic mcparticle_v2::McParticleFlags (getter: flags()), with the idea being that Flags are the ones read in, but the use of the flags getter is captured by the dynamic column, which explicitly includes a protection against the physical primary misclassification. This will mean any derived data creation task that directly asks for flags() will get a preprocessed uint8_t that embeds the correction, meaning a simple re-creation of derived data without any further derived data model change - i.e. even retaining the old mcparticle::IsPhysicalPrimary in the derived tables - should be enough to embed the fix.

Note that this PR is only a draft for now: I'll still test it thoroughly, but wanted to share it here already for discussion. Tagging @jackal1-66, but also @aalkin (with whom I discussed this briefly today, thanks again!) as well as @alcaliva @catalinristea @dsekihat: please also let me know if you have any comments or suggestions. Thank you!