Custom Kernels & First Performance Optimizations

[codecov-commenter]

:warning: Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

Attention: Patch coverage is 57.24816% with 174 lines in your changes missing coverage. Please review.

Project coverage is 76.58%. Comparing base (3fac381) to head (c14ec96).

Files with missing lines	Patch %	Lines
src/qttools/kernels/cuda/dsbcsr.py	0.00%	89 Missing :warning:
src/qttools/kernels/cuda/dsbcoo.py	0.00%	52 Missing :warning:
src/qttools/kernels/cuda/dsbsparse.py	0.00%	16 Missing :warning:
src/qttools/kernels/numba/dsbsparse.py	45.45%	6 Missing :warning:
src/qttools/kernels/__init__.py	54.54%	5 Missing :warning:
src/qttools/datastructures/dsbcoo.py	86.66%	2 Missing :warning:
src/qttools/datastructures/dsbcsr.py	86.66%	2 Missing :warning:
src/qttools/kernels/cuda/__init__.py	0.00%	2 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## dev #58 +/- ## ========================================== - Coverage 84.80% 76.58% -8.22% ========================================== Files 29 37 +8 Lines 1198 1495 +297 ========================================== + Hits 1016 1145 +129 - Misses 182 350 +168 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[vetschn]

I added several custom jitted kernels that should speed up the datastructures quite a bit. Also see the very rough timings over in the comment at #50.

Skipping any kernels not related to the datastructure for now.

Coverage is computed with the Numba jit disabled, but of course the GPU kernels are untested in this pipeline (hence the bad coverage).

Note that the failing test is due to Beyn :skull: