BEANUTILS-539/BEANUTILS-509: 1.X line, switch WeakFastHashMap for ConcurrentWeakHashMap

[car-roll]

Porting the changes made in #56 to the 1.X line

[garydgregory]

This will break binary compatibility and I don't see releasing out of 1.x anyway as we are resource-constrained and want to get 2.0 out.

A general comment: This PR adds an extremely complex class (ConcurrentWeakKeyHashMap) with ZERO tests to match, a no-go.

Since this branch is more than likely going nowhere unless there is a security issue to remedy, I would not spend this on this branch.

[olamy]

@garydgregory this issue makes Jenkins totally irresponsive... see https://issues.jenkins.io/browse/JENKINS-60997 such race condition which depends on the garbage collector (because of weakreference) is definitely impossible to reproduce 100% with a unit test. what could be your proposal alternative to fix that?

[garydgregory]

@garydgregory this issue makes Jenkins totally irresponsive... see https://issues.jenkins.io/browse/JENKINS-60997 such race condition which depends on the garbage collector (because of weakreference) is definitely impossible to reproduce 100% with a unit test. what could be your proposal alternative to fix that?

I expect the new classes to be tested thoroughly. I can understand that the whole use case might be hard to reproduce but a whole new class needs a whole new test to match it. Public and protected APIs cannot be changed.