Parallelize contractions, switch to C++20, and test with GHA

[fevangelista]

This PR does several things:

• Remove dependencies on fmt (now uses std::format)
• Uses #pragma once instead of header guards
• Parallelizes one of the most expensive steps of Wick's theorem (canonicalization)
• Enables testing via GitHub actions

These two changes can speed up the execution of Wick&D significantly. Here are timings for the tests:

C++17 single thread: 127 s (default)
C++17 async:          35 s
C++20 single thread:  68 s
C++20 async:          20 s

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 82.93%. Comparing base (dd7b78e) to head (cafe5b0).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #22      +/-   ##
==========================================
- Coverage   83.91%   82.93%   -0.98%     
==========================================
  Files          47       46       -1     
  Lines        1964     1998      +34     
==========================================
+ Hits         1648     1657       +9     
- Misses        316      341      +25     

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