docs(717): Mode=Unset re-check after c569f806 (no change); a leaving node reports 0

T13b: the isolated otopcua-t13b pair with OpcUa:Redundancy:Mode absent, on origin/master f826af8e
(it contains c569f806) and A/B on c569f806^ (687256d6). Every check passes on both builds:
- levels 250/240;
- Good data on both nodes;
- writes, Acknowledge, AddComment and alarm delivery on the Secondary;
- graceful stop: the survivor reaches 250 259 ms after Leaving (pre-fix 270 ms);
- crash: 250 at +43.6 s (pre-fix +42.8 s), with no demotion at auto-down;
- rejoin at 240 with no failback.
The role-decision changes are gated on IsWarm and do not run under Unset.

Plan L5, the design's section 4.3 and the Warm table in Redundancy.md now say a Leaving/Exiting node
reports ServiceLevel 0, not 90 (ComputeWarm returns 0 for any member that is not Up/Joining; T13 note A).

Claude-Session: https://claude.ai/code/session_01GT67ACmpMR7b9LGRPLijMi

docs/Redundancy.md

@@ -332,7 +332,7 @@ upgrade both nodes before flipping either. `Unset`/`None` keep the unconditional
 
 | Warm state | ServiceLevel |
 |---|---|
-| Down / Detached / endpoint failed | 0 |
+| Leaving / Exiting / Down / Detached / endpoint failed (any member not Up/Joining, so a gracefully leaving Primary drops 250 → 0) | 0 |
 | Boot, before the first computed level | 1 |
 | Standby, stale or DB unreachable | 50 |
 | Standby, healthy | 90 |