Bump to 1.7.1

[h-vetinari]

Not sure why the bot didn't pick this one up yet (and whether I can get further with the build adaptations than in #35)...

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[h-vetinari]

OK, first builds passing again... Now just failing on a (flaky?) test...
FAILED tests/test_residual_quantizer.py::TestAdditiveQuantizerWithLUT::test_RCQ_knn

=================================== FAILURES ===================================
__________________ TestAdditiveQuantizerWithLUT.test_RCQ_knn ___________________

self = <test_residual_quantizer.TestAdditiveQuantizerWithLUT testMethod=test_RCQ_knn>

    def test_RCQ_knn(self):
        ds = datasets.SyntheticDataset(32, 1000, 0, 123)
        xt = ds.get_train()
        xq = ds.get_queries()
    
        # RQ 3+4+5 = 12 bits = 4096 centroids
        rcq = faiss.index_factory(ds.d, "RCQ1x3_1x4_1x5")
        rcq.train(xt)
    
        aq = rcq.rq
    
        cents = rcq.reconstruct_n(0, rcq.ntotal)
    
        sp = faiss.swig_ptr
    
        # test norms computation
    
        norms_ref = (cents ** 2).sum(1)
        norms = np.zeros(1 << aq.tot_bits, dtype="float32")
        aq.compute_centroid_norms(sp(norms))
E        x: array([ 7.76706, 15.41243,  2.32282, ..., 36.52338, 25.93759, 45.54408],
E             dtype=float32)
E        y: array([ 7.76706, 15.41243,  2.32282, ..., 36.52338, 25.93759, 45.54408],
E             dtype=float32)

[h-vetinari]

@conda-forge/vc
It seems that the new compilers in the win-2019 image somehow regressed? The master branch built the same recipe fine essentially, but not this yields weird compilation errors (not related to the version bump, the same errors appear in #45).

I tried downgrading to vs 2017, which worked (mostly - this recipe is just prone to timeouts for win+cuda), see second-to-last CI run...

[h-vetinari]

Though it could also be something with nvcc, because cuda10.2 seems to build fine on vs2019, whereas cuda11.x fails.

CC @conda-forge/nvcc

[h-vetinari]

@conda-forge-admin, please rerender

[h-vetinari]

Some googling again lead to more results this time:

• MSVC 14.29.30037 with CUDA 11.2 occurs error expected a "(" actions/runner-images#3485 ->
• https://developercommunity.visualstudio.com/t/MSVC-C142930037-with-CUDA-112-occur/1437930 ->
• https://developercommunity.visualstudio.com/t/VS-16100-isnt-compatible-with-CUDA-11/1433342 ->
• Non C++17 CUDA compilation is broken when STL headers are included  microsoft/STL#2020

Let's try with C++17 again, though this probably runs back into conda-forge/nvcc-feedstock#51