InfluxDB needs a batch writing instead of a writing by 1 sample

InfluxDB may have a write performance lock when we write a big count of little batches. This writing behavior is our Ceilometer+Gnocchi+InfluxDb writing scenario.

According to the InfluxDb issue description https://github.com/influxdb/influxdb/issues/4492 is better to write a batches with amount of points and series instead of current way.

So, I suggest a implement batch writing for InfluxDB storage driver, what allows to collect samples from our driver and put it by size and time configurable bounds.