FE/Qt: Do not wildcard-disconnect the synthetic work-area workers

[dr1fter]

Fixes #397. Likely also covers #658 and #633, which are the same crash reached
by a shell restart or screen lock changing the host work area.

Problem

r173930 / 700da74 introduced this teardown for the UIInvisibleWindow helpers
that measure host-screen available geometry:

pWorker->disconnect();
pWorker->deleteLater();

The no-argument disconnect() is a wildcard: it also removes Qt's internal
destroyed() connections. Qt warns about exactly this:

QObject::disconnect: wildcard call disconnects from destroyed signal of
UIInvisibleWindow::unnamed

With that bookkeeping gone, any host work-area change can crash the frontend
while the replacement worker is being created:

QAccessibleWidget::text(QAccessible::Text) const
QWidgetPrivate::setWindowTitle_sys(QString const&)
QWidgetPrivate::setWindowModified_helper()
QWidgetPrivate::create()
QWidget::showMaximized()
UIDesktopWidgetWatchdog::sltHandleHostScreenWorkAreaResized(QRect const&)
QGuiApplicationPrivate::processScreenGeometryChange(...)

Notably this needs no monitor hotplug, suspend/resume or DPMS transition. On
X11 an ordinary _NET_WORKAREA change is enough — toggling a desktop panel's
autohide reproduces it on the first attempt, which is why reports involving
shell restarts (#658) and screen locks (#633) look unrelated but are not.

QT_ACCESSIBILITY=0 is sometimes suggested as a workaround; it has no effect
on Qt 6, where the AT-SPI bridge is brought up over D-Bus regardless.

Change

Disconnect only the signal actually connected here —
UIInvisibleWindow::sigHostScreenAvailableGeometryCalculated — at all three
worker-lifecycle sites (sltHandleHostScreenAvailableGeometryCalculated,
updateHostScreenAvailableGeometry, cleanupExistingWorkers), keeping the
deleteLater() from r173930.

Validation

Arch Linux, X11, Cinnamon 6.6.9/Muffin 6.6.3, VirtualBox 7.2.16, Qt 6.11.2.
Trigger: toggling panel autohide, moving _NET_WORKAREA between 1170 and 1200.

	stock	patched
forced work-area cycles	crash on cycle 1	6 cycles, no crash
new coredumps	2	0
wildcard-disconnect warning	present	absent

Note on the wider design

The deeper issue is that the work-area handler itself calls
updateHostScreenAvailableGeometry(), which creates a real window and
maximises it — so measuring the work area perturbs the work area. This change
makes that survivable; a re-entrancy guard in
sltHandleHostScreenWorkAreaResized() would address the shape of it more
directly. I kept to the smaller change because it is the one I could verify.

[oracle-contributor-agreement]

Thank you for your pull request and welcome to our community! To contribute, please sign the Oracle Contributor Agreement (OCA).
The following contributors of this PR have not signed the OCA:

• PR author: dr1fter
• ccwienk@dr1fter.de (@dr1fter)

To sign the OCA, please create an Oracle account and sign the OCA in Oracle's Contributor Agreement Application.

When signing the OCA, please provide your GitHub username. After signing the OCA and getting an OCA approval from Oracle, this PR will be automatically updated.

If you are an Oracle employee, please make sure that you are a member of the main Oracle GitHub organization, and your membership in this organization is public.