DataFrame constructor silently ignores dtype keyword when values cannot be coerced

[TomAugspurger]

In [1]: import pandas as pd

In [2]: pd.DataFrame({'A': ['a', 'b']}, dtype=int)
Out[2]:
   A
0  a
1  b

In [3]: pd.DataFrame({'A': ['a', 'b']}, dtype=int).dtypes
Out[3]:
A    object
dtype: object

I propose that we warn that that will raise in the future.