Refactor: FourMomentum interface

[szabo137]

Using traits to define the Lorentz vector interface seems to limit the implementation just to cartesian coordinates. One solution could be a simplified interface, maybe just for four-momenta. A prototype for such a interface is given in https://github.com/szabo137/FourMomentumBase.jl/pull/4

[szabo137]

Update: As discussed in several issues and offline, the interface for Lorentz vectors especially four-momenta needs a facelift. Therefore, I propose to refactor the interface entirely.

Suggested solution

As mentioned above, in https://github.com/szabo137/FourMomentumBase.jl I implemented a possible interface for four-momenta, which makes use of Julia capabilities for generic implementations, and provides the possibility to register any four-momentum type for a given coordinate system. Once registered, a set of common accessor functions are implemented for the respective momentum type. This should unify the API for four-momenta (and probably Lorentz vectors as well).

I suggest leaving the four-momentum interface in a separate package but moving it from my private account to a more appropriate place. The JuliaHEP organization could be suitable. However, since QEDbase should provide every interface within the QED.jl ecosystem, I propose to make FourMomentumBase a dependency of QEDbase.jl and reexport all types and functions. This avoids too many imports in the downstream packages.

The actual implementation of four-momenta (e.g. SFourMomentum) should be moved to QEDcore.

[szabo137]

Related issues: QEDjl-project/QEDbase.jl#24 #93