design test for non-HMC inversion pathway

[kostrzewa]

We need a test for the non-HMC inversion pathway on both CPU and GPU. This should be based on, say, a single 8c16 gauge config to be stored in the repo in the doc/sample-input directory.

• write input files for a selection of solvers (native, QPhiX, DDalphaAMG, QUDA) with a selection of operators
• run invert in the CI workflow and use the residual check as the test

[Finkenrath]

If useful, we have stored single test configs (L=16,24,32,40,48) under:
https://b2share.eudat.eu/records/1sfke-67w21

[kostrzewa]

Very good, I just think 16c32 or 24c48 might be too large for the github CI. Would be good on daint/beverin, however.