v5.0.1: Windows QtQuick crash fix + node/Tor lifecycle overhaul

Windows crash chain (exposed by the embedded-node default):
- main.cpp: force the QtQuick software renderer. The Mining tab's QQuickWidget
  drove the default RHI/GL path and faulted (0xc0000005) on broken-GL hosts
  (VirtualBox VMSVGA), crashing the app and orphaning the embedded node.
  Real hardware was unaffected.
- ChildProcessGuard + DaemonManager/TorManager: kernel-level die-with-parent
  (Windows Job Object KILL_ON_JOB_CLOSE / Linux PR_SET_PDEATHSIG) so the
  embedded wownerod and bundled tor never survive a wowlet exit or crash.
- DaemonManager: verify-before-adopt via get_info (nettype/height) instead of
  trusting a bare open RPC port, so a stale/orphan node is never adopted.

Node/Tor lifecycle + remote-node connectivity:
- config: torPrivacyLevel default -> allTorExceptNode. NEVER sync blocks over
  Tor; the wallet's node connection is always clearnet. Tor is only for tx
  broadcast, so remote nodes work with or without Tor.
- nodes: add Nodes::reconnect() (clear failure blocklist + force reconnect);
  node/Tor toggle handlers use it so a transient outage no longer strands the
  wallet at "Disconnected".
- MainWindow/SettingsDialog: reconnect on node/Tor toggle and on the async
  daemonStateChanged (closes the start() race; a freshly-toggled local node
  connects immediately).
- TorManager: reset retry/unpack state in init() so a system<->bundled Tor
  re-toggle can start again; quiesce on stop(). Linux unpackBins() rework
  (mkpath + presence-based unpack + graceful return-false), getVersion() arg
  fix, and a missing-binary start guard.
- build.yml: derive TOR_VERSION from the embedded tor binary (a hardcoded
  bundle number made unpackBins rewrite Tor every launch).
- depends/funcs.mk: confine cross-cmake find_package to the depends sysroot so
  host -dev libraries don't leak into the build.

src/utils/ChildProcessGuard.cpp

@@ -0,0 +1,115 @@
+// SPDX-License-Identifier: BSD-3-Clause
+// SPDX-FileCopyrightText: The Monero Project
+
+#include "utils/ChildProcessGuard.h"
+
+#include <QDebug>
+
+#if defined(Q_OS_WIN)
+#include <windows.h>
+#elif defined(Q_OS_UNIX)
+#include <csignal>
+#if defined(Q_OS_LINUX)
+#include <sys/prctl.h>
+#include <unistd.h>
+#endif
+#endif
+
+namespace ChildProcessGuard
+{
+
+#if defined(Q_OS_WIN)
+
+// One process-wide job for all wowlet children. Created lazily on first use; the HANDLE is
+// intentionally never closed — it must stay open for wowlet's entire lifetime so the job persists.
+// On process death (clean OR crash) the kernel closes it, fires KILL_ON_JOB_CLOSE, and reaps the kids.
+static HANDLE g_job = nullptr;
+
+static HANDLE jobHandle()
+{
+    if (g_job)
+        return g_job;
+
+    HANDLE job = CreateJobObjectW(nullptr, nullptr);
+    if (!job) {
+        qWarning() << "ChildProcessGuard: CreateJobObject failed:" << GetLastError();
+        return nullptr;
+    }
+
+    JOBOBJECT_EXTENDED_LIMIT_INFORMATION info{};
+    info.BasicLimitInformation.LimitFlags = JOB_OBJECT_LIMIT_KILL_ON_JOB_CLOSE;
+    if (!SetInformationJobObject(job, JobObjectExtendedLimitInformation, &info, sizeof(info))) {
+        qWarning() << "ChildProcessGuard: SetInformationJobObject failed:" << GetLastError();
+        CloseHandle(job);
+        return nullptr;
+    }
+
+    g_job = job;
+    return g_job;
+}
+
+void installPreStart(QProcess *) { /* no-op on Windows: assignment happens post-start */ }
+
+void adoptAfterStart(QProcess *process)
+{
+    HANDLE job = jobHandle();
+    if (!job)
+        return;
+
+    // QProcess::start() is asynchronous: right after it returns the child often has no PID yet. Wait
+    // (bounded) for it to actually start, else the assign is silently skipped and the child survives a
+    // parent crash (orphaned) — this is exactly what stranded wownerod in testing. Returns the instant
+    // the process is created, so the UI-thread cost is tiny.
+    if (process->state() == QProcess::Starting)
+        process->waitForStarted(5000);
+
+    const qint64 pid = process->processId();   // 0 if it still isn't running
+    if (pid == 0) {
+        qWarning() << "ChildProcessGuard: child has no PID; not assigned to job";
+        return;
+    }
+
+    // AssignProcessToJobObject needs PROCESS_SET_QUOTA | PROCESS_TERMINATE. QProcess does not expose
+    // the child HANDLE, so re-open it by PID.
+    HANDLE proc = OpenProcess(PROCESS_SET_QUOTA | PROCESS_TERMINATE, FALSE,
+                              static_cast<DWORD>(pid));
+    if (!proc) {
+        qWarning() << "ChildProcessGuard: OpenProcess failed for pid" << pid << ":" << GetLastError();
+        return;
+    }
+
+    if (!AssignProcessToJobObject(job, proc))
+        qWarning() << "ChildProcessGuard: AssignProcessToJobObject failed:" << GetLastError();
+    else
+        qInfo() << "ChildProcessGuard: bound child pid" << pid << "to kill-on-close job";
+
+    CloseHandle(proc);   // closing this per-child handle does NOT remove the process from the job
+}
+
+#elif defined(Q_OS_UNIX)
+
+void installPreStart(QProcess *process)
+{
+#if defined(Q_OS_LINUX)
+    // Runs in the forked child between fork() and exec(). Keep it async-signal-safe.
+    process->setChildProcessModifier([] {
+        ::prctl(PR_SET_PDEATHSIG, SIGKILL);
+        // Fork-race guard: if the parent already died before prctl ran, self-terminate.
+        if (::getppid() == 1)
+            ::raise(SIGKILL);
+    });
+#else
+    Q_UNUSED(process)   // macOS/BSD: no PR_SET_PDEATHSIG; rely on aboutToQuit + destructor stop()
+#endif
+}
+
+void adoptAfterStart(QProcess *) { /* no-op on Unix: handled pre-start */ }
+
+#else
+
+void installPreStart(QProcess *) {}
+void adoptAfterStart(QProcess *) {}
+
+#endif
+
+} // namespace ChildProcessGuard