Avoid full repository scans for hotfile replication jobs

[Copilot]

Hotfile replication jobs were being created through the generic migration startup path, which performs a repository-wide initialization scan before the job can leave INITIALIZING. On large pools, that made single-file hotfile jobs unnecessarily expensive and could leave some of them stuck in that startup state.

• Hotfile startup path

  • Change hotfile-triggered jobs to start from the already-resolved CacheEntry instead of invoking the generic full-repository initialization path.
  • This keeps hotfile replication scoped to the single file that crossed the threshold.

• Job initialization API

  • Add a Job.start(Iterable<CacheEntry>) overload for callers that already know the exact entries to enqueue.
  • Preserve the existing asynchronous start() behavior for normal migration jobs that still need a repository scan.

• Initialization handling

  • Factor the shared initialization/cleanup bookkeeping so both startup paths use the same state-transition rules.
  • Keep listener registration and initial enqueueing aligned under the job lock for the hotfile-specific path.

• Regression coverage

  • Add a focused test to verify hotfile startup does not iterate the repository and still creates the expected job.

Example of the behavioral change:

// before
job.start();

// after
job.start(Collections.singletonList(cacheEntry));

[greenc-FNAL]

This PR has been built successfully with tests on a local node; CI in-progress.

[Copilot]

Pull request overview

This PR optimizes hotfile-triggered migration jobs to avoid the generic startup path that scans the entire pool repository, keeping hotfile replication scoped to the specific file that crossed the threshold.

Changes:

• Switch hotfile-triggered jobs to start from an already-resolved CacheEntry (no full repository scan).
• Add Job.start(Iterable<CacheEntry>) to support synchronous startup with preselected entries, while keeping the existing async start() behavior for scan-based jobs.
• Add a regression test asserting hotfile startup does not invoke Repository.iterator().

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
modules/dcache/src/test/java/org/dcache/pool/migration/MigrationModuleTest.java	Adds a regression test to ensure hotfile job startup does not iterate the repository.
modules/dcache/src/main/java/org/dcache/pool/migration/MigrationModule.java	Uses the new Job.start(Iterable<CacheEntry>) overload for hotfile-triggered jobs.
modules/dcache/src/main/java/org/dcache/pool/migration/Job.java	Factors initialization bookkeeping and introduces a synchronous start path that enqueues known entries without scanning.