Prepare threshold computation ready for event generation ("threshold combined mode")

[Juergen Reuter]

Taken over from gitlab issue #87:

Currently, we are generating events for the matched threshold computation by having a process that corresponds to (matched-NLO) and a process that corresponds to NLO. This leads to major efficiency problems when matched and NLO have overlapping distributions such that (matched-NLO) has to sample distributions that go smoothly from positive to negative, which VAMP does not like. Thus, we would prefer to have one master process that directly corresponds to matched to avoid these problems. This also is a benefit for the user in terms of easier handling.

We propose to duplicate the terms in each component (Born, Real and Virtual). Hereby, one term will correspond to the prc_threshold and the other one to the selected ME provider (usually openloops). The result of both terms will be added in each component. This will likely need a new steering option like ?threshold_combined_me_method ? (Open for better names)

Comment by CW:

This is too cumbersome for an ad-hoc implementation. Too many objects which are different for the continuum and the threshold computation are global, especially region_data. The whole issue should be combined with a larger refactoring of the code, which addresses general factorized computations and the merging of the core_manager with the prclib_stack.

[Juergen Reuter]

Transferred back to Gitlab Issue 87.