http pipelining for storageBackend

[julianseeger]

follow up on #577

using HTTP pipelining should reduce that number to essentially 1. (pipelining POSTs is safe for this application imho)
https://hc.apache.org/httpcomponents-core-ga/httpcore-nio/examples/org/apache/http/examples/nio/PipeliningHttpClient.java

[thechauffeur]

See Qabel/qabel-block#64 for another approach maybe.

[enkore]

The two are somewhat orthogonal to each other, although there is a bit of an overlap, because frequent polling will require more open connections if other requests are done at the same time.

What pipelining would do for us is that many small requests (I assume this would happen frequently when syncing directories with many small/average sized files) are accelerated, because they require fewer connections and new requests can be send before responses for earlier requests are received. Since there is no chunking going on long-running requests or responses will still block a connection, so one would need maybe two / just a couple (one+ for smaller requests, one+ for larger requests that shouldn't be pipelined).

I suspect that this may require larger changes overall.