Reuse SeaTunnel containers across selected E2E classes

Per-class engine-container startup and teardown repeatedly paid Docker and cluster-initialization costs during E2E testing.

Adds a JUnit root resource per container type/JVM, serializes class use, and restarts after startup or cleanup failures. Cleans active jobs, volumes, temporary configs and connector JAR additions between classes. Flink and Spark retain their existing lifecycle.

• A warmed local two-class comparison dropped from 12m29s to 11m14s: 75 seconds, approximately 10 percent. This is a slice-specific measurement, not a full-CI speedup claim.
• Reuses engine startup while preserving isolation through cleanup and failure-triggered restarts.
• This is internal test infrastructure, not a production connector change; the broader reuse umbrella remains open.

• The PR reports eight representative E2E tests passing across Flink 1.18/1.20, Spark 3.3 and SeaTunnel Engine, plus focused lifecycle regressions.
• Observed one engine container reused across both classes with job, volume, config and JAR baselines restored. GitHub Build passed and the verified merge commit credits Goutam Adwant.