Floats in known int columns not handled

[calpaterson]

Description

"1.0" should be converted to an int, if the column is known to be integer

Steps to reproduce

1. Create a table with an int column, a
2. PUT a csv file into that table, with 1.0 in the a column

Expected result

1 is stored for that row, in that column

Actual result

500

Additional details

Pandas by default converts int columns to float as soon as a null is added to them, so this is likely to be a very common problem.

Not sure how to deal with 1.000001 and other, similar float issues.