feat(ui): surface daemon DEGRADED mode + v1.3.0 auto-shield status

Rec3 — v1.3.0+ opens a wallet that lost its hdchain in DEGRADED mode (existing funds spendable,
but no new HD-key derivation) instead of aborting. Add a daemon-log classifier
(walletOpenedDegraded) + detectWalletDegraded(), warned once per session. Pre-1.3.0 daemons
never emit that line, so it's a no-op there.

O1 — probe z_autoshieldstatus once per connection (now decoupled from our own toggle/balance) and,
in Settings, show whether the node handles auto-shield itself (+ its destination, or the daemon's
disabled_reason). The checkbox now governs only the wallet's fallback shielder, which defers to the
node. Nothing renders on pre-1.3.0 daemons (no such RPC), so behaviour is unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

src/daemon/daemon_startup_diagnosis.h

@@ -49,6 +49,16 @@ inline bool walletAutoRecovered(const std::string& out)
             && out.find(".bak") != std::string::npos);                             // Recover moved wallet.dat aside
 }
 
+// True when dragonxd (v1.3.0+) opened the wallet in DEGRADED mode: a wallet.dat that lost its hdchain
+// record (e.g. an old `-salvagewallet` output) now OPENS — existing keys stay intact and spendable —
+// instead of aborting, but the daemon can no longer derive NEW HD keys, so z_getnewaddress /
+// z_shieldcoinbase / a t->z z_sendmany fail with "HD seed not found". The only signal is a startup log
+// line; pre-1.3.0 daemons never emit it, so this classifier is naturally a no-op against them.
+inline bool walletOpenedDegraded(const std::string& out)
+{
+    return out.find("Wallet opened in DEGRADED mode") != std::string::npos;
+}
+
 // If `name` is a daemon salvage backup "wallet.<unixtime>.bak", return its timestamp; else -1.
 inline long long parseWalletSalvageBakTs(const std::string& name)
 {