Fix occasional flakiness in the opt_perforce_sdp_backup.service test check (LocalInstallSampleDepotDeepCfgChanges and BackupRecoveryMultiInstance groups).
Root cause: 'systemctl start' on a Type=oneshot unit blocks until the ExecStart process exits, but journald's indexing of that process's final stdout line can lag slightly behind, so the immediately-following 'journalctl -u ... | grep SUCCESS' check occasionally ran before the line was queryable yet -- a pure test-timing race, not a bug in opt_perforce_sdp_backup.sh itself (confirmed pre-existing, seen historically against the SDP Classic fleet too, not tied to any one OS/platform).
Fix: inserted a new synchronization step between 'start the service' and 'confirm success in the journal' that polls (up to 15s, 1s intervals) for the expected string to become queryable before proceeding, using process substitution (<(...)) instead of a literal pipe since cli_tests.cfg's own '|' field delimiter can't appear inside a Command value. Always exits 0 (via a trailing '; true') since its job is just to wait -- the real pass/fail assertion remains entirely in the existing, unmodified 'journalctl' check line that follows it.