fix(swarm): resolve 16 bugs from deep review of ADR-062

CRITICAL:
- Delete stale nvs_provision.bin before provisioning each node
- Fix log filename mismatch: swarm_health.py now finds qemu_node{i}.log
  with node_{i}.log fallback
- CI swarm-test job builds firmware instead of downloading missing artifact
- Accept both qemu_flash.bin and qemu_flash_base.bin as base image

HIGH:
- Replace broad "heap" substring match with precise regex patterns
  (HEAP_ERROR, heap_caps_alloc.*failed, etc.) to avoid false positives
- Guard os.geteuid() with hasattr for Windows compatibility
- Offset SLIRP ports by +100 to avoid collision with aggregator on 5005
- Assertions now WARN (not vacuous PASS) when no parseable data found

MEDIUM:
- Mark network_partitioned_recovery as "(future)" in ADR-062
- Fix node_id prefix dedup bug (node_1 no longer matches node_10)
- Add duplication note in qemu_swarm.py pointing to swarm_health.py
- Document implicit TDM auto-assignment in ADR YAML schema
- swarm_health.py only checks sensor nodes for frame production
- Fix channel 0 treated as falsy

Co-Authored-By: claude-flow <ruv@ruv.net>

docs/adr/ADR-062-qemu-swarm-configurator.md

@@ -92,7 +92,7 @@ nodes:
     node_id: 1
     scenario: 2           # walking person
     channel: 6
-    tdm_slot: 1
+    tdm_slot: 1           # TDM slot index (auto-assigned from node position if omitted)
 
   - role: sensor
     node_id: 2
@@ -141,7 +141,7 @@ assertions:
 | `frame_rate_above` | Each node produces at least N frames/second |
 | `max_boot_time_s` | All nodes boot within N seconds |
 | `no_heap_errors` | No OOM or heap corruption in any log |
-| `network_partitioned_recovery` | After deliberate partition, nodes resume communication |
+| `network_partitioned_recovery` | After deliberate partition, nodes resume communication (future) |
 
 ### Preset Configurations