Use itertools.chain instead of sum to concatenate lists

[Stigjb]

The Python docs discourage sum for concatenating lists, and suggests itertools.chain instead^[1]. I think this makes it clearer that the value is a list and not a number.

Also, the unit test for the evaluate_word_analogies function seemed to have been written for the deprecated accuracy function, where Word2Vec accuracy and KeyedVector accuracy may have been separate methods (?). It seemed pointless to compute analogies twice on the same model, so I changed the test to instead check important aspects of the return value from evaluate_word_analogies.

[menshikh-iv]

Hello @Stigjb, thanks for the PR.

Can you do the same changes for all sum(..., []) please, here are examples that I found (maybe somewhere else?)

gensim/models/keyedvectors.py:1147:            'correct': sum((s['correct'] for s in sections), []),
gensim/models/keyedvectors.py:1148:            'incorrect': sum((s['incorrect'] for s in sections), []),
gensim/models/keyedvectors.py:1253:            'correct': sum((s['correct'] for s in sections), []),
gensim/models/keyedvectors.py:1254:            'incorrect': sum((s['incorrect'] for s in sections), []),
gensim/models/wrappers/ldamallet.py:225:                tokens = sum(([self.id2word[tokenid]] * int(cnt) for tokenid, cnt in doc), [])
gensim/models/wrappers/ldamallet.py:227:                tokens = sum(([str(tokenid)] * int(cnt) for tokenid, cnt in doc), [])
gensim/test/test_corpora_dictionary.py:261:        all_tokens = sum(texts, [])

[menshikh-iv]

Thanks @Stigjb 👍 You are very active last time 🔥 take part in https://hacktoberfest.digitalocean.com/ (free T-shirts)

If you have time and want to help - have a look to #2059, I'll be really appreciated!