This is the companion to #3091 same tests, but on develop (4.1.0-dev)

I observed that negative distance now properly fails instead of silently returning nothing like on main. Tested on PostgreSQL 17.9,
Ubuntu 24.04.

distance = -1.0:

multi-source with distance = -1.0:

The part I wasn't sure about is distance = 0. It still quietly
returns just the origin node on both branches, no error, no warning.

distance = 0.0:

Is that intentional? Mathematically it makes sense (you can only
reach yourself with zero distance), but it feels inconsistent if negative values now raise an error while zero doesn't.

Before I update any docs I wanted to check:
Should distance = 0 stay valid, or get rejected alongside negatives?
Is `XX000 the final error code or still being decided?