gitpush-prestage-abort-test.sh: absorb_policy_handshake's branch has not run since gitpush.sh v5.0, warning 7x per run into nothing

Found while hardening the suite's bare-rm sites for OPS-188, by forcing the new reset assertion red and watching it NOT fire. absorb_policy_handshake() in gitpush-prestage-abort-test.sh keys its then-branch on the output line 'Re-run gitpush.sh after reviewing this info'. gitpush.sh v5.0 removed the policy handshake, so that line no longer exists and the branch has never run since. MEASURED: 7 'WARN: handshake did not fire in <dir> - later assertions may be measuring it' lines on stderr per run, one per make_repo call. The suite still reports 17 passed, 0 failed, because nothing asserts on the WARN. Two costs, both silent: 1. The function invokes the SUT seven extra times per run, each writing a handshake-*.out nobody reads, for an effect that cannot occur. 2. Its own warning says later assertions may be measuring the handshake run. That warning is now unconditionally true and unconditionally ignored - a stderr line that fires on every single run is indistinguishable from a decoration. NOT a false green today: each test gets a fresh repo dir from make_repo, so the .hook-ran markers at :194 and :211 cannot carry across tests, and those two assertions pass on their own hook's marker. The exposure is latent - both assertions PASS when the marker EXISTS, so any future change that reuses a repo dir turns a stale marker into a green line. Decision needed, which is why this is a row and not a commit: delete absorb_policy_handshake and its seven call sites, or retarget it at whatever v5.x handshake replaced the v4 one (if any). Deleting a fixture branch changes what the suite covers. The /bin/rm hardening of its reset shipped with OPS-188 as insurance and is explicitly documented there as not fixing anything observed. NOTE: this suite is one of the six whose SUBJECT is the retiring monolith (OPS-278 leg 3), so the answer may be "it retires with gitpush.sh" - decide this row with that one.