Retire LMAX Disruptor

[simondelabici]

Reasons:

• I have not been able to measure any performance benefit. Premature Optimization ? A BlockingQueue is more flexible and in the CA usage scenario offers comparable or better performance.
• the old disruptor implementation uses a thread per channel monitor and is therefore not scalable.
• it considerably complicates the design creating a significant cognitive learning curve for any new developers who come along.
• adds extra library dependencies.
• it seems hard to get it to shut down cleanly without leaving dangling threads.