fix(table): don't crash on --query lists mixing scalars and objects

[devteamaegis]

What's broken

--output table exits 255 on a --query multi-select list that mixes a scalar with an object. --output json handles the same data fine.

$ aws ec2 describe-instances --output table \
    --query '[Reservations[0].Instances[0].InstanceId, Reservations[0].Instances[0].State]'
ValueError: Row should have 1 elements, instead it has 2

When the column counts happen to line up it's quieter but worse — the object's keys are printed as a data row and the values are dropped:

$ aws ... --output table --query '[[a, b], State]'
+----+-----+
|  a |  b  |
| Code | Name |     <- the 16 / "running" values are gone
+----+-----+

Why it happens

_build_table decided how to treat a list from current[0] alone. If element 0 isn't a dict, every element falls to the row branch — and a dict passes all(self._scalar_type(el) for el in item) there, because that iterates the dict's keys. It's then added as a row of keys, which either loses the values or trips Section.add_row's column-count check.

The fix

Route on whether every element is a dict rather than just the first, and give a dict inside a mixed list a section of its own (the same thing _build_sub_table_from_list already does per element). Homogeneous lists of dicts and lists of lists take exactly the path they did before.

|  i-123            |
+-------------------+
|   OperationName   |
+-------+-----------+
| Code  |   Name    |
+-------+-----------+
|  16   |  running  |

The test

test_mixed_list_with_leading_scalar and test_mixed_list_with_leading_dict in tests/unit/output/test_table_formatter.py. Both fail on develop and pass with the fix:

$ python -m pytest tests/unit/output/test_table_formatter.py -k mixed_list   # before
E   ValueError: Row should have 1 elements, instead it has 2
2 failed, 9 deselected

$ python -m pytest tests/unit                                                # after
2876 passed, 1 skipped