Pull request #1784: PIPE-2858 correct cases where interpolation should use linearpd
Merge in PIPE/pipeline from PIPE-2858-correct-cases-where-interpolation-should-use-linearpd to main
* commit '2e3e82f9f64f7560373a1066f1265f9e750d0b23': PIPE-2858 modify calapp for b2b offset in hifa_diffgaincal to use lienarPD when spws are combined PIPE-2858 add linearPD to almaphcorbandpass apply for combine=spw
Pull request #1834: PIPE-3094 new mstransform based hanning smoothing breaks backward compatibility of vla restores
Merge in PIPE/pipeline from PIPE-3094-new-mstransform-based-hanning-smoothing-breaks-backward-compatibility-of-vla-restores to main
* commit '1353b188830624daeab9438c0396c7181c27aad9': PIPE-3094: Fix hanning_test and remove duplicate test with broken mock setup PIPE-3094: Minior codestyle updates PIPE-3094: Replace field NAME with PHASE_DIR matching PIPE-3094: Bump the minimum pytest version and configure tmp_path retention in `pyproject.toml` PIPE-3094: fix VLA restore backward compat for mstools->mstransform Hanning switching PIPE-3094: Add the `transfer_flagversion` utility function for post-mstransform flag version remapping
PIPE-3094: Fix hanning_test and remove duplicate test with broken mock setup
Two tests (test_file_operation_failure_cleanup,
test_recovery_when_original_ms_removed) were near-identical and both
broken: MagicMock(spec=HanningInputs).keep_original returns a truthy
MagicMock, so the shutil.rmtree branch was never taken and the OSError
never raised, making task_successful always True.
Fix: merge into one test, set keep_original=False explicitly, replace
fragile ordered side_effect list with return_value=True.
