Tests fail locally

[mathurinm]

#43 fixes few failures, but some remain:

========================================================================= short test summary info =========================================================================
FAILED tests/test_linear_regression.py::TestLinearRegression::test_large - ValueError: not enough values to unpack (expected 2, got 1)
FAILED tests/test_linear_regression_large.py::test_weights_vs_sklearn - AttributeError: 'Ridge' object has no attribute 'coef_'
FAILED tests/test_logistic_regression.py::TestLogisticRegression::test_large - ValueError: not enough values to unpack (expected 2, got 1)
FAILED tests/test_logistic_regression_large.py::test_weights_vs_sklearn - AttributeError: 'LogisticRegression' object has no attribute 'coef_'
FAILED tests/test_multiblock.py::TestMultiblock::test_multiblock_fista - AttributeError: 'MultiblockFunctionWrapper' object has no attribute 'at_point'
FAILED tests/test_svd.py::TestSVD::test_fast_sparse_svd - ValueError: not enough values to unpack (expected 2, got 1)
FAILED tests/test_svd.py::TestSVD::test_svd_adjacency_laplacian_coincide - AssertionError: 44.91093483101495 != 44.91093483101497

The "has no attribute coef_" hints at sklearn estimators not being fitted.

Also, the tests take 5 minutes to run. Launching solvers on smaller data may make them faster and thus improve the feedback loop when running the tests.