drivemount: fix use-after-free crash on icon theme change

[Cigydd]

DriveButton connects drive_button_theme_change to the process-global default
GtkIconTheme "changed" signal in drive_button_new() and
drive_button_new_from_mount(), but drive_button_dispose() never disconnected
it. Because the default icon theme outlives the button, after a button is
destroyed (e.g. a volume or mount is removed) a later "changed" emission
invokes the callback on freed memory and schedules drive_button_update() via
g_idle_add(), crashing in the GTK main loop (SIGSEGV inside the
DRIVE_IS_BUTTON() type check).

On affected systems this applet has been crashing repeatedly for months, around
USB device plug/unplug.

Disconnect the handler in drive_button_dispose():

g_signal_handlers_disconnect_by_func (gtk_icon_theme_get_default (),
                                      G_CALLBACK (drive_button_theme_change),
                                      self);

Verified with AddressSanitizer: a harness that destroys a DriveButton and then
emits "changed" on the default icon theme reports heap-use-after-free in
drive_button_queue_update (via drive_button_theme_change) without this
change, and runs cleanly with it.

Fixes #696.

🤖 Generated with Claude Code

[vkareh]

@Cigydd this change looks fine to me, but I would like your commit message attribution to follow MATE's contribution guidelines: https://github.com/mate-desktop/mate-applets?tab=contributing-ov-file#attribution

tl;dr: Clean up the commit message to be more readable, and change the co-authorship like this (or something reasonably similar to it):

- Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
+ Assisted-By: Claude:claude-opus-4.8

[lukefromdc]

I looked over this thankfully very simple code change, basically the same as what a human author would have written. I guess the real question is why wasn't this included when the applet was first written? Or maybe it got lost somewhere?

[lukefromdc]

This builds fine and stops the crash, which I was able to verify by first ejecting a drive with this applet in the panel, then changing icon themes. As expected, that took down the whole panel with all applets in-process.

Applying this change to disconnect the signal handler stops the crash.

Holding off on approval until the attribution is fixed, but if I wrote this code myself the only likely change would be the wording of the comment. I checked the docs for this over at
https://docs.gtk.org/gobject/func.signal_handlers_disconnect_by_func.html