Make `WorkFinder` and the `TransactionRunner`s critical to the Manager

[kevinrr888]

Is your feature request related to a problem? Please describe.
When work on #4524, noticed if the WorkFinder or any of the TransactionRunners die, the Manager will continue to run normally. These are necessary for the Manager to run normally, so the Manager should die if any of these die.

Describe the solution you'd like
WorkFinder and TransactionRunner should be made critical similar to how DeadReservationCleaner is critical to the Manager.

Additional context
#4524 (comment) was the discussion regarding this

[dlmarion]

The TransactionRunner runnables are added to a scheduled executor service. The TransactionRunner can die due to some Error or Exception, but the executor service is monitored and new TransactionRunner's are added as necessary. See:

accumulo/core/src/main/java/org/apache/accumulo/core/fate/Fate.java

Lines 353 to 377 in f52c0b4

	this.fatePoolWatcher =
	ThreadPools.getServerThreadPools().createGeneralScheduledExecutorService(conf);
	ThreadPools.watchCriticalScheduledTask(fatePoolWatcher.schedule(() -> {
	// resize the pool if the property changed
	ThreadPools.resizePool(pool, conf, Property.MANAGER_FATE_THREADPOOL_SIZE);
	// If the pool grew, then ensure that there is a TransactionRunner for each thread
	int needed = conf.getCount(Property.MANAGER_FATE_THREADPOOL_SIZE) - pool.getQueue().size();
	if (needed > 0) {
	for (int i = 0; i < needed; i++) {
	try {
	pool.execute(new TransactionRunner());
	} catch (RejectedExecutionException e) {
	// RejectedExecutionException could be shutting down
	if (pool.isShutdown()) {
	// The exception is expected in this case, no need to spam the logs.
	log.trace("Error adding transaction runner to FaTE executor pool.", e);
	} else {
	// This is bad, FaTE may no longer work!
	log.error("Error adding transaction runner to FaTE executor pool.", e);
	}
	break;
	}
	}
	}
	}, 3, SECONDS));

.

The WorkFinder thread is created via the Threads class, which should be true of almost every server-side thread in the codebase. The Threads created via this class have an UncaughtExceptionHandler set on them that will terminate the VM. See https://github.com/apache/accumulo/blob/main/core/src/main/java/org/apache/accumulo/core/util/threads/AccumuloUncaughtExceptionHandler.java

[kevinrr888]

Ah okay thanks regarding the TransactionRunner. When I get around to this, I'll double check that new runners are created if needed, but the existing code looks good.
When I was checking this, when the WorkFinder died, the termination did not occur... I'll look at this again when I get around to it