Add regression test for aux time tables; correct CHANGELOG entry

The cellstr(curTimeNames) change in split and resample fixes a MATLAB
R2025b failure, not an Octave one: table() rejects a single char name for
'VariableNames', which broke splitting and resampling data whose Aux
holds a table with its own time column (such as a derived HRV series).
Add SplitTest/testAuxTableWithTimeColumn, which fails without the fix,
and describe the bug correctly in the CHANGELOG.

Author: AdrianCurtin

+pf2_base/+tests/+unit/SplitTest.m

@@ -365,5 +365,20 @@ function testSplitDataValuesAreFromOriginal(testCase)
             testCase.verifyEqual(seg.HbO, expectedHbO, 'AbsTol', 1e-15, ...
                 'Extracted HbO should exactly match original data slice');
         end
+
+        function testAuxTableWithTimeColumn(testCase)
+            % An aux signal stored as a table with its own time column
+            % (e.g. a derived HRV series). Splitting and resampling rebuild
+            % its time table, which R2025b's table() rejected when given a
+            % single char variable name.
+            data = testCase.processedData;
+            t = (data.time(1):1:data.time(end))';
+            data.Aux.hrv = table(t, 60 + randn(numel(t), 1), ...
+                'VariableNames', {'time', 'RMSSD'});
+            seg = pf2.data.split(data, data.time(1) + 20, data.time(1) + 80);
+            testCase.verifyTrue(isfield(seg, 'Aux'), 'split dropped the Aux field');
+            rs = pf2.data.resample(data, 0.5);
+            testCase.verifyTrue(isfield(rs, 'Aux'), 'resample dropped the Aux field');
+        end
     end
 end

CHANGELOG.md

@@ -28,7 +28,7 @@ Groundwork for running headless single-subject processing under GNU Octave.
 - `pf2_base.tests.integration.FullPipelineTest/testDifferentMethods` builds its method combinations with the Pipeline API and runs them through a `ProcessingContext`, so it no longer skips on machines without the named methods in their library or changes the global method selection.
 - `pf2_base.normalizeMarkerDict` parses `{code, label}` cell dictionaries element by element, so numeric codes keep one row each under Octave.
 - `pf2_base.external.zp2sos` orders complex roots with `sortrows` instead of `sort(..., 'ComparisonMethod', 'real')`, which Octave lacks; the resulting second-order sections are identical.
-- `pf2.data.resample` and `pf2.data.split` pass aux time-column names to `table` as a cell array, which Octave requires.
+- `pf2.data.split` and `pf2.data.resample` failed on aux signals stored as tables with their own time column (for example a derived HRV series): MATLAB R2025b's `table` rejects a single char name for `'VariableNames'`. The time-column name is now passed as a cell array. v1.1.0 is affected the same way.
 - `pf2.import.importHitachiMES` no longer requests the unused second output of `strsplit`, and `pf2.data.split` assigns its time-size variables directly; both failed under Octave.
 - `loadDeviceCfg` and `pf2_base.gui.updateCurrentDevice` assign whole table columns instead of growing them by indexed assignment, and no longer call `sprintf` on a categorical or `nansum` on a logical.
 - `bvoxy` declares its positional `coefs` and `isOD` options before its name-value parameters. Argument handling under MATLAB is unchanged.