[BUG] ndarray_from_mp_obj() needs inclusive check of negative max values

[sfe-SparkFro]

Describe the bug

micropython-ulab/code/ndarray.c

Lines 1625 to 1629 in 88ef893

	if(ivalue > -128) {
	dtype = NDARRAY_INT8;
	} else {
	dtype = NDARRAY_INT16;
	}

ivalue > -128 should be ivalue >= -128

Similar here:

micropython-ulab/code/ndarray.c

Lines 1617 to 1622 in 88ef893

	if((ivalue < -32767) || (ivalue > 32767)) {
	// the integer value clearly does not fit the ulab integer types, so move on to float
	ndarray = ndarray_new_linear_array(1, NDARRAY_FLOAT);
	mp_float_t *array = (mp_float_t *)ndarray->array;
	array[0] = (mp_float_t)ivalue;
	} else {

ivalue < -32767 should be ivalue < -32768

ulab version: 6.7.4-4D-c

To Reproduce

from ulab import numpy as np

foo = np.ones((1), dtype=np.int8) * -128
print("should be int8:", foo)

foo = np.ones((1), dtype=np.int8) * -32768
print("should be int16:", foo)

Expected behavior

Should print:

should be int8: array([-128], dtype=int8)
should be int16: array([-32768.0], dtype=int16)

Instead prints:

should be int8: array([-128], dtype=int16)
should be int16: array([-32768.0], dtype=float32)

It incorrectly promotes the dtype to int16 and float respectively.

Additional context

Testing on my end, making the suggested changes fixes the problem.

[v923z]

Thanks for finding the root cause of the glitch. The issue is addressed in #724.