Write batching is unaware of transport message size limits

[simonaldrich]

When writing batches of points to Influx (via the InfluxDB::batchOf mechanism) over UDP it is easy to accidentally exceed the maximum UDP packet size. This causes an exception to be raised in UDP::send which ultimately results in the loss of the entire batch of points.

I propose that InfluxDB::flushBatch should be aware of the largest message that can be transmitted over its Transport and sends the Points in the largest batch possible.