Fast-to-sql

Fixes to issuer 13

Added the `copy` parameter which is `False` by default. The `fast_to_sql` function does some altering of the column names of the dataframe passed to it. If `copy` is set to `False`, the original columns will be changed. If it is `True`, a copy of the dataframe will be made local to the `fast_to_sql` function and the original will not be altered. The tradeoff here, is you are using more memory by making a copy. If you are just doing ETL, where your final step is running `fast_to_sql`, and you don't need the dataframe anymore, then keep this set to `False`

Fixes to issue 15

`fast_to_sql` was previously setting the SQL column type of `datetime64[ns]` as `datetime`, but it should have been `datetime2`. With this change null values now correctly upload for a `datetime64[ns]` column.

This release attempts to fix 9 and 12.

Fix for 9:
- Removed the `_clean_str_cols` function - pointed out by simbr80 as not needed
- Added support for `np.NaN` columns - these will show up as `NULL` in SQL now

Fix for 12
- Attempt to fix schema issue by getting default schema from server if one is not provided