Implement a count up latch paradigm with a managed blocker

The class `CountUpLatch` is fine when using normal threads, but doesn't fit with a fork join pool model. It is unwise to ever block in a fork join pool, since you effectively reduce the maximum concurrency by 1 when you do so. We should create a class with similar properties as `CountUpLatch`, except with a managed blocker pattern.

Start by locating the existing CountUpLatch class and its usages, then inspect how fork join pool tasks are handled in the repository. Compare the required managed blocker pattern with CountUpLatch's behavior; done means a similar count-up latch can be used without directly blocking fork join pool workers.

Written by the indexing model from the issue text.