Fix TriggerZoneQuadPoint forwarding heading into the parent's radius slot

[MacrayBlackhand]

TriggerZoneQuadPoint.init forwards its arguments to the parent
TriggerZone.init positionally:

super().__init__(_id, position, hidden, name, color, properties, heading, link_unit_id)

Since PR #254 re-added radius as the parent's 7th positional parameter
_id, position, hidden, name, color, properties, radius, heading, link_unit_id),
this call is now misaligned: heading lands in the radius slot, link_unit_id
lands in heading, and link_unit_id is dropped. A quad zone constructed with a
non-default heading serializes "radius": (and "radius": null at
defaults), whereas the DCS Mission Editor writes "radius": 0 for quad zones.

This isn't covered by #251/#243/#254 -- those fixed TriggerZoneCircular's
forwarding when radius was re-exposed but didn't update TriggerZoneQuadPoint,
so the gap went unnoticed. The existing test_create_quad_point_zone never threads
a heading/link_unit_id through the quad path, so it doesn't exercise this.

Fix: forward to the parent by keyword and pass radius=0 (the DCS-canonical value
for quad zones). Keyword forwarding also prevents this class of positional
regression from recurring.

Adds a round-trip test that constructs a quad zone with an explicit heading and
link_unit_id and asserts radius == 0 with both surviving the save/load.