fix: bulletproof first-boot container creation and install reliability

Remove the Bitcoin RPC 60-second gate that blocked 13+ dependent containers
(mempool, electrumx, btcpay, lnd, fedimint) from being created on first boot.
Containers now always get created and auto-restart via health monitor once
Bitcoin becomes responsive — the designed recovery path.

Additional hardening:
- Validate archy-net creation with retry (silent failure broke DNS)
- Verify critical images are loaded, re-load from tarballs if missing
- Create SearXNG settings.yml before container start (was missing)
- Run reconciler automatically after first-boot failures
- Add load-images as explicit systemd dependency with 900s timeout
- Propagate config write errors in install.rs (bitcoin.conf, lnd.conf)
- FileBrowser password change: retry loop (6 attempts) + 0o600 perms
- Post-start verification: detect containers that exit immediately
- Add 2s dependency waits between container starts

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

scripts/container-specs.sh

@@ -412,11 +412,13 @@ load_spec_searxng() {
   SPEC_IMAGE="${SEARXNG_IMAGE}"
   SPEC_PORTS="8888:8080"
   SPEC_MEMORY="$(mem_limit searxng)"
+  SPEC_VOLUMES="/var/lib/archipelago/searxng:/etc/searxng"
   SPEC_HEALTH_CMD="curl -sf http://localhost:8080/ || exit 1"
   SPEC_READONLY="true"
-  SPEC_TMPFS="/tmp:rw,noexec,nosuid,size=256m /run:rw,noexec,nosuid,size=64m /etc/searxng:rw,noexec,nosuid,size=16m"
+  SPEC_TMPFS="/tmp:rw,noexec,nosuid,size=256m /run:rw,noexec,nosuid,size=64m"
   SPEC_TIER="3"
   SPEC_CAPS=""
+  SPEC_DATA_DIR="/var/lib/archipelago/searxng"
 }
 
 load_spec_onlyoffice() {