572 check for singularity when the solver parameters flag is turned on

[K20shores]

Closes #572

This PR fixes an unknown bug where a zero in the bottom right position of the U matrix (which leads to a singular matrix) was not detected by always checking for zero in the bottom right element. Additionally, two tests are added to check for singular matrices when running rosenbrock

I learned that to get a singular U matrix, at least for these simple 2 and 3 species reactions, we needed a negative reaction rate, which I think is not physically realistic. This, in addition to the knowledge that chemical simulations have been running for decades without a singularity check, seem to indicate that chemical systems are not structured in a way that would lead to singular systems.

[sjsprecious]

Thanks @K20shores for addressing this issue. I guess the CUDA version of LU decomposition (https://github.com/NCAR/micm/blob/main/src/solver/lu_decomposition.cu#L96-L102) needs to be revised accordingly as well? And update the test_cuda_rosenbrock.cpp?

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.47%. Comparing base (6a60a49) to head (03b15ef).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #603      +/-   ##
==========================================
+ Coverage   93.17%   93.47%   +0.30%     
==========================================
  Files          49       49              
  Lines        3500     3494       -6     
==========================================
+ Hits         3261     3266       +5     
+ Misses        239      228      -11     

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

[sjsprecious]

Thanks @K20shores for working on this and figuring out this bug. I could confirm that all the tests passed on Derecho's A100 GPU.

I also talked with Allison today, and she told me that we could construct a singular matrix with linear dependence between two rows or two columns (which is the definition of "singularity"). Not sure whether this is easy to do but this may avoid the usage of negative reaction rates.

[mattldawson]

looks great!

[boulderdaze]

Looks good! Just a couple questions

[mwaxmonsky]

Looks great @K20shores!