Improvement of NCH

[qbarthelemy]

This PR:

• improves NCH code,
• completes test_distance (distance to hull of a point contained in the set is null),
• completes test_classification adding QuanticNCH (forgotten in Implementation of NCH (Nearest Convex Hull) classifier #253),
• improve dosctring of classification module.

[gcattan]

@qbarthelemy Can you complete the light benchmark, and check the scores after and before the changes ?

[toncho11]

Example file classify_P300_nch.py needs to be updated with the new parameter name subsampling and validated that it runs (python errors) and it completes correctly.

[toncho11]

Currently I have this during the tests:

FAILED tests/test_utils_distance.py::test_qdistance_logeuclid_to_convex_hull[optimizer0] - assert 8.567767282575056e-07 == 0 ± 1.0e-12
FAILED tests/test_utils_distance.py::test_qdistance_logeuclid_to_convex_hull[optimizer1] - assert 8.567767282575056e-07 == 0 ± 1.0e-12

[qbarthelemy]

Currently I have this during the tests:

FAILED tests/test_utils_distance.py::test_qdistance_logeuclid_to_convex_hull[optimizer0] - assert 8.567767282575056e-07 == 0 ± 1.0e-12 FAILED tests/test_utils_distance.py::test_qdistance_logeuclid_to_convex_hull[optimizer1] - assert 8.567767282575056e-07 == 0 ± 1.0e-12

Have you update/pull my branch with my last commits?

[toncho11]

Currently I have this during the tests:
FAILED tests/test_utils_distance.py::test_qdistance_logeuclid_to_convex_hull[optimizer0] - assert 8.567767282575056e-07 == 0 ± 1.0e-12 FAILED tests/test_utils_distance.py::test_qdistance_logeuclid_to_convex_hull[optimizer1] - assert 8.567767282575056e-07 == 0 ± 1.0e-12

Have you update/pull my branch with my last commits?

I do:
git pull origin pull/255/head and I see the last changes from @gcattan coming
pip install .[docs] or pip install .[tests]
pytest
and I still get it:

>       assert dist == pytest.approx(0)
E       assert 8.567767282575056e-07 == 0 ± 1.0e-12
E         comparison failed
E         Obtained: 8.567767282575056e-07
E         Expected: 0 ± 1.0e-12

[gcattan]

LGTM! @qbarthelemy @toncho11 thank you for your work!