CRAMS update after the 2-population CR spectra modelling work

[nj-vs-vh]

Main changes:

• Full Python interface to CRAMS. It is generated with SWIG and is wrapped into a custom Python module. It cuts down on evaluation time ~50% w.r.t. the current subprocess & file I/O method in mcmc. The installation is currently only manual, but a pip install crams is achievable in a reasonable amount of time.
• Cosmetic changes to the C++ side to accomodate the new interface. Mainly it's around the Runner object that now encapsulates the logic previously found in main.cpp.
• Physics changes to CRAMS to support the knees project:
  • Cross-sections (both inelastic and fragmentation) are extrapolated logarithmically to high energies to support calcualtion above 10 TeV. They are mostly flat or slowly varying, so the extrapolation should not introduce large errors.
  • Source spectra now support features, currently SBPL and erfc cutoff are implemented.
• Tests are updated where the implementation has minor changes. All existing entrypoints and workflows are preserved, no breaking changes introduced.

UPD:

• source spectra can have multiple features; the ones currently implemented are break and erfc cutoff parametrized with population variance of Rmax, more to be added