Stabilize nightly cluster stop/resume integration test timing

[Copilot]

The nightly integration suite was failing intermittently in ClusterScenarioIT.stopResumeCompletes because the stop signal could occur after the first run had effectively completed, causing resume assertions to compare against a new crawl session. This change tightens the scenario so stop/resume is exercised during true partial progress.

• Failure mode addressed

  • stopResumeCompletes expected session continuity on resume, but occasionally observed a new crawlSessionId when first run progressed too far before stop.

• Targeted test hardening

  • Increased first-run mock fetch delay in this scenario (100ms → 250ms) to widen the interruption window.
  • Tightened partial-progress gate before issuing stop:
    • from “some progress but not complete”
    • to “some progress and still below half of total refs”

• Scope

  • Single-file test-only change in:
    • crawler/core/src/test/java/com/norconex/crawler/core/cluster/ClusterScenarioIT.java

baseConfig(numOfRefs, 250).accept(cfg);

return imported > 0
        && imported < (numOfRefs / 2);