The sync test job was attempting to import the remote V2 organization
immediately after the CADT health check, but the wallet was still
syncing. Add a wallet sync polling step (up to 20 min) before the
CADT API readiness check and org import.

- Set DATALAYER_FILE_SERVER_URL to null instead of localhost to fully
  prevent mirror creation (AUTO_MIRROR_EXTERNAL_STORES=false alone
  does not prevent mirrors created during store subscription)
- Reorder startup: wait for wallet sync BEFORE starting CADT so
  background tasks don't fail continuously on wallet sync errors
- Pre-subscribe to org store via chia RPC and wait for datalayer to
  sync it before calling the CADT import API
- Add wallet re-sync wait after subscriptions since new DL subs
  cause wallet to re-derive puzzle hashes
- Add retry loop for org import since importOrganization returns
  early (without creating DB record) if the store is not yet synced

The governance node has separate store IDs for V1 and V2:
- V1: 29fe490b... (main store with only a 'v1' key)
- V2: 4212425c... (main store with a 'v2' key)

The sync test was using the V1 governance body ID for the V2 config,
so GovernanceV2.sync() could never find the v2 sub-store and the
orgList/pickList never synced.

The V2 API returns arrays directly, not {data: [...]} objects like V1.
The jq selector '.data[] // .[]' fails on arrays because '.data' on
an array is an error, not null. Use a conditional expression that
handles both formats.

The wallet may report synced=true at the Chia RPC level while still
being unavailable or not fully ready for datalayer store transactions
at the CADT layer. This is especially likely when a real
GOVERNANCE_LOOKUP_ID is configured, since CADT subscribes to the
governance store immediately on startup, triggering additional wallet
activity.

Three-layered fix:
- Add waitForWalletReadyForTransactions helper that polls /v1/health
  for 3 consecutive successes before proceeding
- Call this helper in the upgrade test's before() hook to confirm
  wallet stability before attempting org creation
- Add outer retry loop (3 attempts, 90s settle delay) around the
  entire create + waitForV1OrganizationReady cycle so that a
  PENDING-org-disappeared failure (wallet not fully synced for
  transactions) retries the whole creation rather than failing
- Extend CI wait step to poll CADT /v1/health after chia rpc
  reports synced=true, requiring 3 consecutive 200s before tests run