mgr/influx: Use Queue to store points which need to be written

This allows us to multiplex data being send to Influx as we have
a configurable amount of workers sending data to Influx.

The main bottleneck for the performance seems to be fetching all
the perf counters using this code:

    self.get_all_perf_counters()

On a larger cluster, for example 2000 OSDs this can take about 20s
where flushing to Influx only takes 5s.

A 2000 OSD cluster generates about 100k data points on every run,
prior to using a Queue these would all be send to Influx in series
in that took over 15 seconds to complete.

Python Six is being used in the code to make sure it's compatible
with both Python 2 and 3.

Signed-off-by: Wido den Hollander <[email protected]>

doc/mgr/influx.rst

@@ -50,11 +50,13 @@ For example, a typical configuration might look like this:
     
 Additional optional configuration settings are:
 
-:interval: Time between reports to InfluxDB.  Default 5 seconds.
+:interval: Time between reports to InfluxDB.  Default 30 seconds.
 :database: InfluxDB database name.  Default "ceph".  You will need to create this database and grant write privileges to the configured username or the username must have admin privileges to create it.  
 :port: InfluxDB server port.  Default 8086
 :ssl: Use https connection for InfluxDB server. Use "true" or "false". Default false
 :verify_ssl: Verify https cert for InfluxDB server. Use "true" or "false". Default true
+:threads: How many worker threads should be spawned for sending data to InfluxDB. Default is 5
+:batch_size: How big batches of data points should be when sending to InfluxDB. Default is 5000
 
 ---------
 Debugging