Fix some of the warnings/ deprecated functions

[FredHappyface]

Closes #3066
Addresses #3014

Remaining (removed windows specific warnings):

.tox\py39-win\lib\site-packages\scipy\sparse\sparsetools.py:21
  c:\users\dell\documents\github\gensim\.tox\py39-win\lib\site-packages\scipy\sparse\sparsetools.py:21: DeprecationWarning: `parse.sparsetools` is deprecated!
  scipy.sparse.sparsetools is a private module for scipy.sparse, and should not be used.
    _deprecated()

gensim/test/test_api.py::TestApi::test_load_dataset
gensim/test/test_api.py::TestApi::test_multipart_load
  c:\users\dell\documents\github\gensim\.tox\py39-win\lib\site-packages\smart_open\smart_open_lib.py:479: DeprecationWarning:unction is deprecated.  See https://github.com/RaRe-Technologies/smart_open/blob/develop/MIGRATING_FROM_OLDER_VERSIONS.rst fo
information
    warnings.warn(message, category=DeprecationWarning)

gensim/test/test_ldaseqmodel.py::TestLdaSeq::test_doc_topic
gensim/test/test_ldaseqmodel.py::TestLdaSeq::test_dtype_backward_compatibility
gensim/test/test_ldaseqmodel.py::TestLdaSeq::test_topic_word
  C:\Users\Dell\Documents\GitHub\gensim\gensim\models\ldaseqmodel.py:297: RuntimeWarning: divide by zero encountered in doublrs
    convergence = np.fabs((bound - old_bound) / old_bound)

gensim/test/test_lsimodel.py::TestLsiModel::test_online_transform
gensim/test/test_lsimodel.py::TestLsiModel::test_online_transform
gensim/test/test_lsimodel.py::TestLsiModel::test_online_transform
  c:\users\dell\documents\github\gensim\.tox\py39-win\lib\site-packages\numpy\matrixlib\defmatrix.py:1109: PendingDeprecation: the matrix subclass is not the recommended way to represent matrices or deal with linear algebra (see https://docs.scipy.orumpy/user/numpy-for-matlab-users.html). Please adjust your code to use regular ndarray.
    return matrix(concatenate(arr_rows, axis=0))

gensim/test/test_miislita.py: 1 warning
gensim/test/test_sharded_corpus.py: 15 warnings
  C:\Users\Dell\Documents\GitHub\gensim\gensim\interfaces.py:92: UserWarning: corpus.save() stores only the (tiny) iteration 
in memory; to serialize the actual corpus content, use e.g. MmCorpus.serialize(corpus)
    warnings.warn(

[mpenkov]

Is there any real benefit in doing this instead of:

Suggested change

	import collections.abc as collections
	import collections.abc

You're saving 4 characters, but making it less obvious what's actually being used.

[FredHappyface]

I was thinking of keeping small diffs at the time but I agree that it would make more sense to change it

[mpenkov]

Yes, that is a somewhat valid reason, but think about the trade-off:

1. We do a bit more work during this review (one-off cost)
2. People reading the code may be confused (repetitive cost)

Better to pay once up front and save future generations of gensim developers the pain (I'm exaggerating, obviously, but you probably get my point).

[FredHappyface]

Yeah absolutely! I'll get that changed. Hopefully I'll have it sorted in a couple hours

[mpenkov]

OK, merging. Thank you for your work @FredHappyface !