This PR fixes Issue #1914, where scalar values (int, float, complex, bool, np.number) were not correctly coerced into NdArray or JaxArray. PR adds a complete test suite ensuring correct behavior for all scalar types

Root Cause:

np.ndarray(value) and jnp.ndarray(value) interpret numeric values as shapes, not as data. This resulted in:
-uninitialized arrays when value was an integer
-validation errors when value was a float
-no coercion for complex or boolean scalars

1. Updated NdArray validator

elif isinstance(value, (int, float, complex, bool, np.number)):
    arr_from_scalar = np.array([value])
    return cls._docarray_from_native(arr_from_scalar)

2. Updated JaxArray validator

elif isinstance(value, (int, float, complex, bool, np.number)):
    arr_from_scalar = jnp.array([value])
    return cls._docarray_from_native(arr_from_scalar)

3. New Tests added in test_ndarray.py

test_from_scalar_int
test_from_scalar_float
test_from_scalar_complex
test_from_scalar_bool
test_from_scalar_in_document

These tests confirm:
Scalars convert to (1,) shape, Values are preserved correctly , Dtypes are correct (int32/int64, float32/float64, complex, bool) ,Scalars work correctly when assigned as Document fields