Ensure JobClusterActor's logs always contain the clusterName

[sileshidev-lab]

Summary

Fixes #385 — log lines emitted from JobClusterActor didn't consistently include the cluster name, making it hard to trace a specific cluster's behavior in logs (~256 log call sites in this class, added ad-hoc over time, many without name).

Approach: rather than manually retrofitting every existing call site (tedious, easy to miss, and doesn't guard against new call sites forgetting it later), this overrides aroundReceive — the single method every message this actor processes passes through — to set the cluster name in SLF4J's MDC before delegating to the actual handler, and clear it in a finally block afterward. This makes the cluster name automatically available to every log line emitted anywhere during that message's handling, for any logging backend that renders MDC values (structured/JSON encoders, or a pattern layout with %X{clusterName}).

Verified AbstractActorWithTimers.aroundReceive (via the Timers trait) already delegates further up the chain to AbstractActor.aroundReceive, so calling super.aroundReceive(receive, msg) from the override preserves existing timer bookkeeping — confirmed by inspecting the actual Akka 2.6.15 akka-actor jar's bytecode rather than assuming the API shape.

Test plan

• New test testAroundReceivePutsClusterNameInMdcDuringMessageHandlingAndClearsItAfter in JobClusterAkkaTest. Since MDC is thread-local, it exercises aroundReceive directly via TestActorRef (bypassing the mailbox/dispatcher) with a receive handler that captures the MDC value synchronously during "processing," then asserts it matches the cluster name and is cleared afterward.
• akkaTest task (this class's actual test task — this project splits *AkkaTest classes into a separate Gradle task from test): 58 passed (57 pre-existing + 1 new), 0 regressions
• test task for the module: 358 passed, 0 regressions
• compileJava clean, no new warnings on the modified file (verified against a stashed baseline)