Using matrix market examples for testing

[AndPotap]

Many of the unit tests that we have are simple cases that are far from the ones observed in real-life. With the goal of stress testing our implementations we selected a few matrix market benchmarks that can run in a reasonable time but that are also hard.

• [X] CG
• [X] GMRES
• [X] Lanczos
• [X] Arnoldi
• [X] LOBPCG

[codecov[bot]]

Codecov Report

Attention: 7 lines in your changes are missing coverage. Please review.

Comparison is base (7b9c0ac) 81.95% compared to head (bffc6e8) 81.82%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #74 +/- ## ========================================== - Coverage 81.95% 81.82% -0.13% ========================================== Files 40 40 Lines 3302 3317 +15 ========================================== + Hits 2706 2714 +8 - Misses 596 603 +7 ``` | [Files](https://app.codecov.io/gh/wilson-labs/cola/pull/74?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=wilson-labs) | Coverage Δ | | |---|---|---| | [cola/linalg/decompositions/lanczos.py](https://app.codecov.io/gh/wilson-labs/cola/pull/74?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=wilson-labs#diff-Y29sYS9saW5hbGcvZGVjb21wb3NpdGlvbnMvbGFuY3pvcy5weQ==) | `74.59% <ø> (ø)` | | | [cola/linalg/eig/lobpcg.py](https://app.codecov.io/gh/wilson-labs/cola/pull/74?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=wilson-labs#diff-Y29sYS9saW5hbGcvZWlnL2xvYnBjZy5weQ==) | `100.00% <100.00%> (ø)` | | | [cola/linalg/inverse/cg.py](https://app.codecov.io/gh/wilson-labs/cola/pull/74?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=wilson-labs#diff-Y29sYS9saW5hbGcvaW52ZXJzZS9jZy5weQ==) | `97.36% <ø> (ø)` | | | [cola/backends/jax\_fns.py](https://app.codecov.io/gh/wilson-labs/cola/pull/74?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=wilson-labs#diff-Y29sYS9iYWNrZW5kcy9qYXhfZm5zLnB5) | `95.12% <40.00%> (-1.38%)` | :arrow_down: | | [cola/linalg/eig/eigs.py](https://app.codecov.io/gh/wilson-labs/cola/pull/74?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=wilson-labs#diff-Y29sYS9saW5hbGcvZWlnL2VpZ3MucHk=) | `80.83% <42.85%> (-2.36%)` | :arrow_down: |

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

[AndPotap]

Hey @mfinzi, could you take a look at this PR. Basically, I added some difficult sparse linear operators (from the matrix market) under tests/data and I'm calling them to tests different algorithms: CG, GMRES, Lanczos, Arnoldi and LOBPCG. I'm only enabling the smallest test for the majority of the algorithms as the other cases take hours to run. Maybe this is fine for some night runs, I guess not for when people are doing PRs...