[FIX] : Fix issue #422
with this implementation, now DataFrameClient allow to push DataFrame with null values (NaNs or None)

• Issue: DataframeClient fails to push DataFrame with NaNs
• Root cause : null are not allowed on Influx (null is not a valid type)
• Possible work around: push column per column (with dropnan() before each push)
• Solution : corresponding points are dropped from the list of points pushed to influx. Also takes into account full lines with nulls and DataFrame full of nulls)

[FIX] : Fix issue #479
conversion of DatetimeIndex to epoch was not correct on some machines.

• Issue : on some Windows machines, DatetimeIndex not correctly converted to epoch
• Root cause: to convert to Epoch, timestamp is first converted to int. On some machines, default conversion to int leads to a conversion to int32 (which lead to incorrect conversion to epoch). Conversion to int should be to int64.
• Solution: conversion forced to int64 in the code