macosunifiedloggingreceiver: report component status for permanently-failing polls

[kroepke]

Problem

A permanently-failing poll in macosunifiedloggingreceiver is retried forever while the collector reports itself healthy.

readLogs (receiver/macosunifiedloggingreceiver/receiver.go) logs the error and backs off to max_poll_interval, then keeps retrying. Nothing else observes the failure.

Some failures are deterministic and will never succeed:

• A malformed predicate. log exits 64 (EX_USAGE), e.g. log: Bad predicate (Unable to parse the format string "subsystem =="). The predicate is not validated in-process, by design — log owns its grammar.
• A predicate referring to a field log rejects.

For these, the receiver emits nothing indefinitely and the only signal is a repeating log line. This matters because the predicate is fleet-pushed: a bad value can be deployed to many hosts at once.

Recommendation

Report component status instead of only logging:

• Call componentstatus.ReportStatus with a permanent error when a poll fails with a deterministic cause (log exit 64), so a health-check extension surfaces the receiver as unhealthy to the fleet manager.
• Keep retrying transient failures with the existing backoff and StatusRecoverableError.

Do not fail Start instead: that aborts the whole collector, taking down unrelated pipelines on the host.

Requires confirming the collector build wires componentstatus and a health-check extension.

Notes

Deliberately out of scope: distinguishing every log exit code. Exit 64 is sufficient to separate "config is wrong" from "read failed".