sys: byteorder: Add 40 bit variants

[rettichschnidi]

This adds sys_*_*40 functions that operation on a 40 bits values.

The motivation for this PR is to allow extending the net_buf functionality to support 40 bit values (#70177).

[andyross]

OK, I'll bite: why do we want support for variable-endian 5-byte integers in OS code? This seems like some kind of special purpose data type for a particular app or hardware or problem area? Do any other systems have libraries like this?

Just seems like code that wants to live closer to whatever it's application area is.

[rettichschnidi]

OK, I'll bite: why do we want support for variable-endian 5-byte integers in OS code? This seems like some kind of special purpose data type for a particular app or hardware or problem area?

I want to extend the net_buf functionality to support 40 bit values (#70177). As the existing net_buf functionality makes use of Zephyr kernel code, I figured it makes sense to keep the additional net_buf code aligned and extend the kernel code accordingly.

Do any other systems have libraries like this?

No idea.

Just seems like code that wants to live closer to whatever it's application area is.

I think the discussion boils down to whether one agrees that the 40 bit functionality in net_buf makes sense.

[rettichschnidi]

Marking as draft again, because I need to sort out #70305 first. done

[rettichschnidi]

Just seems like code that wants to live closer to whatever it's application area is.

I think the discussion boils down to whether one agrees that the 40 bit functionality in net_buf makes sense.

Closing this PR because its justification depends entirely on #70177. Better to discuss only once whether adding such functionality to net_buf is a valuable contribution.