fix(math): preserve non-differenced dimension shape in diff for 2D arrays

When diff is applied to a 2D array and the number of differences (n)
exceeds the size of the operated dimension, the function previously
returned allocate(y(0, 0)), incorrectly collapsing both dimensions.

This fix preserves the size of the non-differenced dimension:
- dim=1: allocate(y(0, size(x, 2)))
- dim=2: allocate(y(size(x, 1), 0))

Author: JAi-SATHVIK

src/math/stdlib_math_diff.fypp

@@ -91,7 +91,11 @@ contains
         size_work = size_x + size_prepend + size_append
 
         if (size_work <= n_) then
-            allocate(y(0, 0))
+            if (dim_ == 1) then
+                allocate(y(0, size(x, 2)))
+            else
+                allocate(y(size(x, 1), 0))
+            end if
             return
         end if