Core: Do not retry Tasks after interrupt

[huaxingao]

Tasks.foreach(...).retry(...) treated every failure as retryable unless it was exactly UnrecoverableException. When a job is cancelled, FileIO often wraps InterruptedException in RuntimeException / UncheckedIOException. Iceberg then slept and retried, and the retry could succeed after the caller was already gone (expire-snapshot cleanup is one path that hits this).

This change walks the cause chain. If the failure is InterruptedException, InterruptedIOException, or ClosedByInterruptException, do not retry, restore the interrupt flag, and rethrow.

[pvary]

There are other places as well where we don't handle interrupts correctly. See: #18236