What changed

Add a 30-minute deadline to every unbounded Recycler/Checker polling loop under cloud_p0/recycler.

The change now covers all four unbounded loops across the three related cases:

• test_recycler.groovy: wait for a Recycler round and wait for a Checker round
• check_meta.groovy: wait for a successful Recycler round
• test_checker.groovy: wait for the expected Checker failure

All other Recycler cases already use bounded retry counters and are unchanged. When a service timestamp does not advance, the affected test now fails with the relevant timestamps instead of looping until TeamCity’s 24-hour build timeout.

Why

The branch-4.1 Recycler pipeline has repeatedly remained inside an unbounded polling loop with no test occurrence until TeamCity terminated the build after 1440 minutes (for example, builds 204533 and 204268). A normal successful test_recycler execution in build 203828 took about 11 minutes, so a 30-minute per-phase deadline keeps reasonable headroom while turning an infrastructure/service stall into an actionable failure.

This does not suppress a Recycler or Checker failure and does not change any success condition. It only bounds waits and preserves the last observed job timestamps in the error message.

Validation

• git diff --check
• Scanned regression-test/suites/cloud_p0/recycler: 4 unbounded while (true) loops across 3 files, all 4 now have timeout guards
• All other polling loops in the Recycler directory already have explicit retry bounds
• Compared with successful test_recycler duration in branch-4.1 build 203828
• Current build 204938 has remained in the same unbounded loop for hours with no test occurrence, reproducing the need for this guard

Related: DORIS-27416