build-it 4.1.3

Patch release. Two defects in the checkpoint gate found when the real READINESS of the first real run
was put through it. IDs are that run's findings.

Fixed

• H-22 a program that merges slice by slice could never merge: gate --action merge named the
  adversarial and pre-merge checkpoints, which cover every wave and can only run after the last one.
  plan-it 5.2.0 now writes no pre-merge when the plan's merge_model is incremental, and the
  closing reviews gate the release instead. The reader accepts such a block: the "no pre-merge
  checkpoint" check asks for a pre-release one when inputs.merge_model is incremental.
• H-27 due --at merge returned only the pre-merge checkpoint while gate --action merge also
  needed the adversarial one. due --at merge and due --at release now list every checkpoint whose
  gates the matching gate --action checks (release includes the promotion ones, as the gate does).

Note

• The 4.1.2 entry below labels the due --at merge|release prefix-match fix as H-22. The run's id
  for that finding is H-25; H-22 is the merge lock fixed here. The test names
  H-22: due --at ... cover the H-25 fix.