Correlator not able to handle long paths on Windows.

The correlator has been enhanced to support loading EPL monitor files via the YAML configuration file with paths that are longer than MAX_PATH on Windows.

Correlator deadlocks if some EPL is runnable while other EPL causes a timeout of a stop-the-world-request.

In some circumstances, a correlator request which needs to suspend contexts (such as toStringAll) but which timed out due to a long-running monitor execution could leave other contexts suspended after it times out, leading to other operations deadlocking. This issue has now been fixed.

Concurrent, first request to Prometheus endpoint caused a correlator crash.

Concurrent accesses to the /metrics endpoint on the correlator could result in a crash, and also when new /metrics statistics were first reported, they held the value zero (first access only). The risk of crashing is now removed, and the current value, of any new statistics, will be retrieved when accessed for the first time.

First Prometheus metrics request returns ...memory_bytes 0.000.

Access to new /metrics statistics would report the value zero (first access only). The current value, of any new statistics, will now be retrieved when accessed for the first time.