Remove scikit-learn by implementing precision_recall_fscore_support

[mgoin]

Smoke test:

import numpy

y_true = numpy.array(["cat", "dog", "pig", "cat", "dog", "pig"])
y_pred = numpy.array(["cat", "pig", "dog", "cat", "cat", "dog"])

print("numpy", precision_recall_fscore_support_np(y_true, y_pred))

from sklearn.metrics import precision_recall_fscore_support
print("sklearn", precision_recall_fscore_support(y_true, y_pred))

output:

numpy (array([0.66666667, 0.        , 0.        ]), array([1., 0., 0.]), array([0.8, 0. , 0. ]), array([2., 2., 2.]))
sklearn (array([0.66666667, 0.        , 0.        ]), array([1., 0., 0.]), array([0.8, 0. , 0. ]), array([2, 2, 2]))