Use TestSuite for QR and LQ

[kshyatt]

Goal here is to break up #93 into more manageable diffs, starting with LQ and QR

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/implementations/lq.jl	50.00%	1 Missing ⚠️

Files with missing lines	Coverage Δ
ext/MatrixAlgebraKitGenericLinearAlgebraExt.jl	95.65% <100.00%> (ø)
src/implementations/qr.jl	91.38% <100.00%> (-5.24%)	⬇️
src/implementations/lq.jl	95.18% <50.00%> (-3.75%)	⬇️

... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[lkdvos]

This looks great!

I think the main comment would just be about still testing the different algs as well, since I think we are now not being that thorough about this.

With the idea of having this be useful for downstream tests as well, how do you feel about introducing a general function with the following signature in the test suite, which can then be overloaded if need be as well:

test_function(::typeof(qr_full!), T::Type, sz; kwargs...)
test_function(::typeof(qr_full!), T::Type, sz, alg::AbstractAlgorithm; kwargs...)

where the first signature is the one we already have, which uses keyword arguments to find an algorithm automatically, and the second is a different signature for testing out a provided algorithm instead?

---

Finally, I was wondering if it makes sense to intersperse the GPU tests with the CPU ones like that, or to simply keep a little bit of boiler plate and still have them in separate files.
Part of this is because I would like to eventually move to https://github.com/JuliaTesting/ParallelTestRunner.jl, which automatically detects tests by files.

[kshyatt]

The GPU tests are currently in the same files, AFAICT?

[github-actions]

Your PR no longer requires formatting changes. Thank you for your contribution!

[kshyatt]

OK, added some algorithm tests to the QR test suite as an example, if we like this then I'll do the same to LQ. The idea is to just cover every supported algorithm for each array backend and make sure some basic properties work. We don't have a good way to check ispivoted(alg), for example, so the tests are quite barebones.

[lkdvos]

Looks great to me, I think only the remaining comment about testing the positive-ness of the alg implementation if need be is what I would still add, otherwise this is good to go for me.

Thanks for all the work on this, I really think this is a massive improvement!

[kshyatt]

I think only the remaining comment about testing the positive-ness of the alg implementation if need be is what I would still add, otherwise this is good to go for me.

The LQ tests still need their version of the algorithm tests!

[kshyatt]

OK, now the LQ algo tests are in!

[kshyatt]

OK, the GenericLinearAlgebra stuff is now tested as well. I had to add some special casing because qr_null just doesn't work there, and it's a big weird that we did support Float16 for Diagonal but not generic arrays (when it would work with GLA)

[Jutho]

Two minor comments, but otherwise looks great!

[lkdvos]

Except for the minor comment about being able to locally run the /gputests, I'd be happy to get this merged!

[kshyatt]

OK, I think that is everything!