There is a race condition otherwise between release_gdi_font and
find_cached_gdi_font, leading to invalid memory access:

One thread calling release_gdi_font may decrement refcount to 0,
then try to enter font_lock. At the same time, another thread may be
calling find_cached_gdi_font through select_font, holding the font_lock.

This second thread would find refcount set to 0, and then try to remove
unused_entry from its list, although it hasn't been added yet to the
unused list.
Signed-off-by: Rémi Bernon <rbernon@codeweavers.com>
Signed-off-by: Huw Davies <huw@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>