Add FuzzTargetsCount index and error log for bad targets

[dylanjew]

When the fuzz_task_output has an invalid fuzz target, FuzzTargetCount.count = 0. This can happen when there are no fuzz target binaries found, or the build is entirely broken.

When this happens, no more fuzz tasks would be scheduled. We recover from this scenario by falling back to the default modifier when the FuzzerJob.multiplier would be 0:

clusterfuzz/src/clusterfuzz/_internal/cron/fuzzer_and_job_weights.py

Lines 380 to 383 in f547870

	# If the count is 0, it may be due to a bad build or some other issue. Use
	# the default weight in that case to allow for recovery.
	if targets_count and targets_count.count:
	multiplier = targets_count.count

This is a health signal, and we should monitor it properly. This PR adds an index so we can look these up, and a specific error we can track this with.

@aakallam, I'm not sure whether we want to set up a metric for an alert on this, because we're going to have very large dimensionality here and we can just scan the logs/Datastore for bad fuzz targets. But I'm open to just throwing a Monarch metric here if that's the easiest way to alert us that something is broken.

Related to b/537750310

Testing

I pushed this change onto dev

[aakallam]

Starting with an alert and then removing it if it is too noisy sounds good to me

[dylanjew]

Starting with an alert and then removing it if it is too noisy sounds good to me

I'll see if we can set up alerting based on the log, otherwise I'll just send another PR to add a Monarch metric if that turns out to be difficult.