[BUG] cudf.concat resets field for struct columns

[shaneding]

Describe the bug
When concatenating two struct columns, the resulting struct field is reset.

Steps/Code to reproduce bug

>>> import cudf
>>> sr1 = cudf.Series([{"a": 5}, {"c": "hello"}, {"b": 7}])
>>> sr2 = cudf.Series([{"a": 5, "c": "hello", "b": 7}])
>>> cudf.concat([sr1, sr2])
0        {'0': 5.0, '1': None, '2': None}
1    {'0': None, '1': None, '2': 'hello'}
2        {'0': None, '1': 7.0, '2': None}
0      {'0': 5.0, '1': 7.0, '2': 'hello'}
dtype: struct
>>> 

Expected behavior
The result of the concatenation should be of the form

0        {'a': 5.0, 'b': None, 'c': None}
1    {'a': None, 'b': None, 'c': 'hello'}
2        {'a': None, 'b': 7.0, 'c': None}
0      {'a': 5.0, 'b': 7.0, 'c': 'hello'}

[shaneding]

I have narrowed the issue to the fact that concat in reshape.py is not properly re-populating the fields for the struct. However, one issue that needs to be addressed is when the two series to be concatenated have different fields for their structColumn, in that case we have no real way of generating fields in the correct final series. Hence, I am thinking of throwing an error for this particular case when there is a field mismatch.