[ENH]: Sensitvitiy Analyzer for sklearns SGD classifier

[adswa]

This PR contains a number of changes all aimed at providing a Sensitivity Analyzer for an SDG classifier.
It...

• implements the subclass SKLLearnerAdapterWeights(Sensitivity) to get the feature weights from sklearns SGDClassifier, wrapped with the SKLLearnerAdapter
• implements the subclass MulticlassClassifierSensitivity(BoostedClassifierSensitivityAnalyzer) to store the labels of the pairwise comparisons in a SampleAttribute of the weights Dataset.
• implements the necessary get_sensitivity_analyzer() functions in clfs/skl/base.py and clfs/meta.py
• adds the classifier to the warehouse

Current problems:

• the solution to checking whether the classifier in question wrapped with SKLLearnerAdapter() is indeed an SGD is crude: if 'SGDClassifier' in str(self._skl_learner): This is because I couldn't find a way to get the information on the type of classifier before training. Probably there is a prettier solution.
• 2 tests fail:

======================================================================
FAIL: mvpa2.tests.test_hdf5_clf.test_h5py_clfs
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/adina/env/USA/local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/adina/env/USA/local/lib/python2.7/site-packages/nose/util.py", line 620, in newfunc
    return func(*arg, **kw)
  File "/home/adina/Repos/PyMVPA/mvpa2/testing/tools.py", line 291, in newfunc
    func(*(arg + (filename,)), **kw)
  File "/home/adina/Repos/PyMVPA/mvpa2/tests/test_hdf5_clf.py", line 115, in test_h5py_clfs
    cmp_(error, error_)
  File "/home/adina/env/USA/local/lib/python2.7/site-packages/numpy/testing/_private/utils.py", line 865, in assert_array_equal
    verbose=verbose, header='Arrays are not equal')
  File "/home/adina/env/USA/local/lib/python2.7/site-packages/numpy/testing/_private/utils.py", line 789, in assert_array_compare
    raise AssertionError(msg)
AssertionError: 
 Single scenario lead to failures of unittest test_h5py_clfs:
  on
    lrn=<MulticlassClassifier> :
     
Arrays are not equal

(mismatch 100.0%)
 x: array([0.022222])
 y: array([0.033333])


======================================================================
FAIL: Some really basic testing for match_distribution
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/adina/Repos/PyMVPA/mvpa2/tests/test_stats_sp.py", line 211, in test_match_distribution
    self.assertTrue('norm' in names)
AssertionError: False is not true

----------------------------------------------------------------------

• two tests raise errors:

======================================================================
ERROR: Test analyzers in split classifier
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/adina/Repos/PyMVPA/mvpa2/testing/sweep.py", line 69, in do_sweep
    method(*args_, **kwargs_)
  File "/home/adina/Repos/PyMVPA/mvpa2/tests/test_datameasure.py", line 119, in test_analyzer_with_split_classifier
    sens = sana(ds)
  File "/home/adina/Repos/PyMVPA/mvpa2/base/learner.py", line 258, in __call__
    return super(Learner, self).__call__(ds)
  File "/home/adina/Repos/PyMVPA/mvpa2/base/node.py", line 137, in __call__
    result = self._call(ds, **(_call_kwargs or self._get_call_kwargs(ds)))
  File "/home/adina/Repos/PyMVPA/mvpa2/measures/base.py", line 1053, in _call
    return self.__combined_analyzer._call(dataset)
  File "/home/adina/Repos/PyMVPA/mvpa2/measures/base.py", line 909, in _call
    sensitivity = analyzer(dataset)
  File "/home/adina/Repos/PyMVPA/mvpa2/base/learner.py", line 258, in __call__
    return super(Learner, self).__call__(ds)
  File "/home/adina/Repos/PyMVPA/mvpa2/base/node.py", line 137, in __call__
    result = self._call(ds, **(_call_kwargs or self._get_call_kwargs(ds)))
  File "/home/adina/Repos/PyMVPA/mvpa2/clfs/meta.py", line 1214, in _call
    senses.sa[clf.get_space()] = [(clf.clfs[i].poslabels[0], clf.clfs[i].neglabels[0]) for i in range(len(clf.clfs))]
  File "/home/adina/Repos/PyMVPA/mvpa2/base/collections.py", line 599, in __setitem__
    str(self)))
ValueError: Collectable 'targets' with length [6] does not match the required length [12] of collection '<SampleAttributesCollection: biases,lrn_index,targets>'.

======================================================================
ERROR: mvpa2.tests.test_emp_null.test_efdr
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/adina/env/USA/local/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/adina/Repos/PyMVPA/mvpa2/testing/tools.py", line 344, in newfunc
    return func(*arg, **kwargs)
  File "/home/adina/Repos/PyMVPA/mvpa2/tests/test_emp_null.py", line 36, in test_efdr
    np.testing.assert_array_less(efdr.fdr(2.9), 0.15)
  File "/home/adina/Repos/PyMVPA/mvpa2/support/_emp_null.py", line 332, in fdr
    self.learn()
  File "/home/adina/Repos/PyMVPA/mvpa2/support/_emp_null.py", line 245, in learn
    medge = medge[whist]
IndexError: boolean index did not match indexed array along dimension 0; dimension is 133 but corresponding boolean dimension is 132

[adswa]

Forgot to tag you, @yarikoptic