Fix context propagation in tomcat thread pool

[laurit]

Resolves #4497
Tomcat ThreadPoolExecutor extends java.util.concurrent.ThreadPoolExecutor and overrides the execute method. It also uses a custom task queue that reports queue full when pool size is smaller than max. This makes j.u.c.ThreadPoolExecutor create new threads. If maximum number of threads is reached then j.u.c.ThreadPoolExecutor will reject the task because it could not be queued (queue behaved as it is full) and thread could not be created. Tomcat ThreadPoolExecutor catches RejectedExecutionException and places task in queue. The problem is that we clear propagated context when j.u.c.ThreadPoolExecutor.execute fails with RejectedExecutionException because we interpret an exception in execute as a failure to queue task and assume that it could never execute.
This pr alters context propagation handling so that when task already has expected context we just skip it and return null, which disables the cleanup on exception logic.
Hopefully this pr also gets rid of debug logs Failed to propagate context because previous propagated context is where new and old context are the same and have different debug locations.

[trask]

💯