The decision-number ceiling drops to thirteen

Flagged "growing debt" in a reassessment and was wrong about the important half. The duplicate decision numbers are already capped: ./orbistoun.sh check fails on any new one, and fails again if a listed number stops duplicating, so docs/decision-number-backlog.txt is a ratchet rather than a backlog. It cannot grow, and the stated reason for not batch-fixing is sound - each one means renumbering an entry and its citations, so "take one when you are already in that area".

One of the fourteen was mine. I wrote ## D157 and later ## D157 (resolved), using the same number for a superseding entry because it read as a continuation. It reads as a duplicate, and someone else had to absorb it into the ceiling.

Taken, being about as in-that-area as it gets. The resolving entry is D178; all three citations referred to the original, which keeps D157. Ceiling: 14 -> 13.

Two of those citations were also stale - the knowledge file and a code comment both still said the direct-memory mapping was parked and off, which stopped being true when it was turned on by default. Fixed while in there, which is the same principle: a citation is only worth having if what it points at is still true.

The lesson for me: a "(resolved)" entry under the same number is not a superseding pattern, it is a collision. Supersede with a new number that names what it resolves.