StatefulTreeCache uses the string representation of GUID as an element in an Fqn. Problem is GUID.toString() is very slow. We need to optimize this

One alternative is to use the GUID itself in the Fqn instead of the string. But, a test showed that a serialized GUID is 10x as big as the String (over 400 bytes bigger). And serializing/deserializing a GUID was ~ 5x more time consuming than the String. Hence the original use of the String.

So, we need to optimize.

One thing is to fix get() where it ends up calling GUID.toString() twice.

Second is to find a way to cache the string so it gets created once per SFSB rather than once per get, release and replicate call.

JBCOMMON-88 is to do that in the GUID class itself. If we want to get this into EJB3 quicker than a new JBCOMMON release would allow we can evaluate using an org.ejb3 GUID subclass that caches the toString() result.