Fix bug for Mol2VecFingerprint

[nissy-dev]

I found the bug about the usage of mol2vec function, so I fixed.

[coveralls]

Coverage decreased (-0.04%) to 80.204% when pulling 3ddc610 on nd-02110114:fix-mol2vec into 0fcc6b4 on deepchem:master.

[peastman]

What bug is this fixing?

[nissy-dev]

I made a mistake about how to use sentences2vec method. I should pass the List for the first argument of sentences2vec function, but I didn't. As a result, the return value of sentences2vec was a N×M matrix and I misunderstood that we should apply the some aggregation method like sum or mean to convert the N×M matrix to a M dim vector.

The correct usage is to pass the list to sentences2vec method. By doing so, we can get a vector which we expects, not matrix.

[nissy-dev]

I think this PR is ready for the review

The CI failed by test_weave_fit_simple_infinity_distance, which is not related to this PR.
Recently, the CI is broken by test_weave_fit_simple_infinity_distance

[rbharath]

This looks good from my pass!

@nd-02110114 You know this featurizator much better than I do, so feel free to merge in once you feel this is ready :)

[nissy-dev]

Thanks you for checking! I merge in!