Separate simplification of operators from SummedOp.add and fix bugs of add

[t-imamichi]

Summary

This PR separates same operator check in SummedOp.add and makes another method SummedOp.simplify to summing up same operators.
It also fixes all bugs found in #979 and add unit tests.

Fixes #979

minor change: remove unnecessary [] in all and any.

Details and comments

It fixes following bugs.

• ListOp.equals did not compare ListOp.coeff -> Fixed SummedOp.add is slow due to linear search and some bugs related to coefficients #979 (comment)
• SummedOp.add did not handle SummedOp.coeff correctly, especially when coeff != 1 -> Fixed SummedOp.add is slow due to linear search and some bugs related to coefficients #979 (comment)
• SummedOp.add did not apply simplification in a particular case -> Separate out the simplification as another method and the simplification won't be applied automatically. Note that simplification should not be often invoked because it takes O(n^2) time where n is the number of operators. SummedOp.add is slow due to linear search and some bugs related to coefficients #979 (comment)

[t-imamichi]

I also fixed some comments of AbelianGrouper.

[Cryoris]

LGTM, thanks for restructuring the tests.

[t-imamichi]

Thank you for your feedback!

[Cryoris]

The spellchecker is complaining about the word subops, if you add it to the .pylintdict the CI should succeed.

[t-imamichi]

I changed the comment to be clear.

[woodsp-ibm]

If you looked in the build output then the spell checker did provide alternatives like sub-ops. In general if it can be spelled out or changed its better than adding something like subops, which is not really a word, to the custom dict. So your change was the right thing in my book. Sometimes though its unavoidable as you need to refer to something defined elsewhere.

[t-imamichi]

Travis failed on Test Aqua 1 Python 3.8 due to the following error.

{1} test.aqua.operators.legacy.test_tpb_grouped_weighted_pauli_operator.TestTPBGroupedWeightedPauliOperator.test_evaluate_qasm_mode [2.386731s] ... FAILED
Captured traceback:
~~~~~~~~~~~~~~~~~~~
    Traceback (most recent call last):
      File "/home/travis/build/Qiskit/qiskit-aqua/test/aqua/operators/legacy/test_tpb_grouped_weighted_pauli_operator.py", line 148, in test_evaluate_qasm_mode
    self.assertGreaterEqual(reference[0].real,
      File "/opt/python/3.8.0/lib/python3.8/unittest/case.py", line 1316, in assertGreaterEqual
    self.fail(self._formatMessage(msg, standardMsg))
      File "/opt/python/3.8.0/lib/python3.8/unittest/case.py", line 753, in fail
    raise self.failureException(msg)
    AssertionError: 1.736805965121753 not greater than or equal to 1.7490732614333155

[t-imamichi]

I'm wondering why results varied even if the seed of simulator is fixed.

[dongreenberg]

This is a great addition, but I think it would probably be more natural to put the new simplify functionality inside SummedOp's reduce, rather than create a new function (unless simplify is slow).