Skip slow calculation of reference C6

[marvinfriede]

The loops in the _load_c6 function are actually really slow and create a bottleneck for the whole calculation. Since the output of this function is not dependent on the system, we only use it to store the tensor (".pt" file) and now skip the calculation and directly load the tensor from the "reference-c6.pt" file.

This gives a speed-up of more than one order of magnitude for small systems.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (583c5a1) 99.48% compared to head (d48be67) 99.48%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #39      +/-   ##
==========================================
- Coverage   99.48%   99.48%   -0.01%     
==========================================
  Files          20       20              
  Lines         391      390       -1     
  Branches       43       44       +1     
==========================================
- Hits          389      388       -1     
  Misses          1        1              
  Partials        1        1              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[awvwgk]

We can also replace the npy file with the one that is already unpacked from the composite indices rather than storing a packed and an unpacked Version in two different files next to each other.

[awvwgk]

LGTM