Fix combnk MATLAB behavior

[sahilphad07-sudo]

combnk did not match MATLAB behavior for non-positive values of k.

For k = 0, Octave returned a 0x0 empty result instead of MATLAB's 1x0 empty row vector. Negative values of k were also rejected, while MATLAB returns a 1x0 empty row vector.

This change updates the handling of non-positive k values to match MATLAB behavior.

Before

octave:> combnk(1:3, 0)
ans =

[](0x0)

octave:> combnk(1:3, -1)
error: combnk: second input argument must be a non-negative integer

After

octave:> combnk(1:3, 0)
ans =

[](1x0)

octave:> combnk(1:3, -1)
ans =

[](1x0)

MATLAB Behavior

>> combnk(1:3, 0)
ans =

     1×0 empty double row vector

>> combnk(1:3, -1)
ans =

     1×0 empty double row vector

Regression Tests Added

• k = 0
• Negative k

Test Results

All existing and new combnk BISTs pass:

PASSES 6 out of 6 tests.

[pr0m1th3as]

Thank you, but I am closing this without merging.

The k = 0 part is right: MATLAB returns a 1-by-0 result of the input's class, one empty combination, and I will apply that myself. A negative k has no meaning, though, and MATLAB returning 1x0 for it, even for k = -2.5, is not behaviour worth copying, so combnk keeps refusing it. As written, the patch also accepts negative values while its error message still asks for a non-negative integer.