CoreMessageReceiver.returnMesagesLoopDaemon is being executed too often causing higher CPU consumption

## Actual Behavior
CoreMessageReceiver.returnMesagesLoopDaemon is scheduled to be executed every millisecond even if there are neither prefetched messages nor pending receivers, causing unnecessary higher CPU usage.

## Expected Behavior
Solution to this could be to use blocking queue and its `take()` method or `poll(time, unit)` instead of `poll()` for `prefetchedMessages` and `pendingReceives` collections.

## Versions
- OS platform and version: Windows 10 enterprise
- Maven package version or commit ID: 3.2.0 (related code looks the same as in version 3.4.0)