Insert of Pandas DataFrame loses nanoseconds

[genzgd]

If inserting a Pandas Timestamp object with nanosecond precision into a ClickHouse column with datatype DateTime(9) (which should have nanosecond precision), the nanosecond value is lost. The root cause is that the underlying transformation to a ClickHouse type for Pandas Timestamp includes a conversion to a Python datetime.datetime, which only has microsecond precision.