Refactoring of Probabilistic Linear Solvers

[JonathanWenger]

Summary

Introduces an abstract base class for probabilistic numerical methods and refactors all variants of the probabilistic linear solver to a compositional pattern.

Content

Refactoring of PLS

• Refactors probabilistic linear solvers to a compositional pattern. Closes Refactor the design of the probabilistic linear solver functions and classes #51 and closes Refactor probnum.linalg to resolve the pylint messages #158.
• Better testing of uncertainty calibration. Closes Test uncertainty calibration procedure for probabilistic linear solvers more thoroughly #100
• Different types of prior belief initializations. Closes Probabilistic Linear Solver should accept deterministic quantities as prior arguments #109

Other issues which might be tackled in this PR are #174.

Other

[codecov]

Codecov Report

Merging #283 (2fe886a) into master (ddada48) will decrease coverage by 5.17%.
The diff coverage is 83.31%.

❗ Current head 2fe886a differs from pull request most recent head ebaba02. Consider uploading reports for the commit ebaba02 to get more accurate results

@@            Coverage Diff             @@
##           master     #283      +/-   ##
==========================================
- Coverage   84.21%   79.03%   -5.18%     
==========================================
  Files         103       73      -30     
  Lines        5238     3954    -1284     
  Branches      695      457     -238     
==========================================
- Hits         4411     3125    -1286     
- Misses        610      688      +78     
+ Partials      217      141      -76     

Impacted Files	Coverage Δ
src/probnum/diffeq/__init__.py	100.00% <ø> (ø)
src/probnum/filtsmooth/__init__.py	100.00% <ø> (ø)
src/probnum/quad/__init__.py	100.00% <ø> (ø)
src/probnum/random_variables/__init__.py	100.00% <ø> (ø)
src/probnum/linalg/linearsolvers/matrixbased.py	54.69% <10.52%> (-2.55%)	⬇️
...nalg/linearsolvers/_probabilistic_linear_solver.py	43.10% <43.10%> (ø)
src/probnum/utils/arrayutils.py	70.00% <50.00%> (+32.50%)	⬆️
src/probnum/linops/_linear_operator.py	79.20% <64.51%> (-1.78%)	⬇️
src/probnum/linalg/_problinsolve.py	53.68% <66.66%> (ø)
src/probnum/linops/_kronecker.py	65.42% <66.66%> (-20.58%)	⬇️
... and 153 more

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[JonathanWenger]

Parts of this PR are already in the library via the linked PRs. The open todos are tracked in issues.