Keep 0 and False in concat merge outputs - #1323

What / why

Task.concat merged multi-action outputs with a truthiness filter (if y), so a 0, 0.0 or False result was dropped from the merged string instead of being written into it:

task = Task([lambda x: [0, 5], lambda x: ["Widget", "Gadget"]], merge="concat")
list(task([1, 2]))
# before: ['Widget', '5. Gadget']
# after:  ['0. Widget', '5. Gadget']

A quantity of 0 is real data. Only None and "" contribute nothing but a separator, so the guard now skips exactly those - the same rule the tabular pipeline already applies in Tabular.concat (column is not None and column != "", covered by testFalsyValues).

Closes #1322

Checks

test/python/testworkflow.py on Python 3.11, base 893d042:

• new testMergeWorkflowFalsyValues fails on base with AssertionError: Lists differ: ['Widget', '5. Gadget'] != ['0. Widget', '5. Gadget'] and passes with this change
• merge coverage stays green: testMergeWorkflow, testMergeUnbalancedWorkflow and testIteratorActions (which loops over hstack/vstack/concat/None)
• whole module: 6 failed, 35 passed with this change vs 6 failed, 34 passed on a base checkout of the same environment - the 6 failures are the identical set of pre-existing local gaps (optional pipeline/workflow extras not installed, image model), none is an assertion in the merge path
• black --check clean, pylint 10.00/10 with the repo .pylintrc
• numpy/torch path checked directly: np.stack rows give scalars, and np.float64(0.0) != "" returns np.bool_ without warnings under -W error::DeprecationWarning, so array merges keep working (['0.0. 1.0', '5.0. 6.0'])

Note

AI-assisted: this change was drafted and verified locally with an AI coding assistant (reproduction, RED/GREEN runs and lint above).