[dheerchheda77]

Follow-up to #522. crosstab and tabulate refused datetime and duration vectors with their own type checks before reaching grp2idx, while MATLAB accepts both in both functions.

Changes

• crosstab: datetime and duration inputs take the existing path for text and categorical inputs, through grp2idx, so labels are the group names and missing values (NaT, NaN) are not counted. A datetime or duration matrix is refused with the existing "must be vectors" error.
• tabulate: datetime and duration are added to the accepted types and to the error message. A new branch drops missing values with ismissing and numbers the groups with grp2idx, in sorted order, returning the same 3-column cell table as the other non-numeric types.
• Help text of both functions lists the new types; tabulate also states that NaT and NaN durations are ignored and that the values are sorted.
• doc-cache entries regenerated for both functions.

Comparison with MATLAB R2026b

Data: three dates repeated six times with one NaT, against 1, 2 and 3 hours.

Call	MATLAB R2026b and Octave
[t, chi2, p, labels] = crosstab (d, h)	t = [0 6 0; 5 0 0; 0 0 6], chi2 = 34, p = 7.4519e-07, labels '01-Mar-2025' ... and '1 hr' ...
tabulate (d)	sorted dates, counts 6, 5, 6, percents 35.2941, 29.4118, 35.2941 (NaT left out of the total)
tabulate (h)	'1 hr', '2 hr', '3 hr', counts 6, 6, 6, percents 33.3333

Testing

Octave 11.3.0, datatypes 1.5.0:

• crosstab: 34 of 34 pass, 3 tests new.
• tabulate: 35 of 35 pass, 9 tests new. Edge cases cover a NaN duration, row input, all-NaT and empty input, printed output, and matrix input.
• grp2idx and knnsearch (the other caller of crosstab) still pass.

Not included

Other functions where MATLAB R2026b refuses datetime are left alone: dummyvar, mnrfit, and fitctree / TreeBagger / fitcensemble.