Reduce overload on filesystem events (alternative patch)

When a big number of files are added to, or removed from, the desktop or the trash, the system generates a big number of filesystem events. With the current code, the current operation of getting the disk info is cancelled and a new operation is launched, all this for each event received. This poses a big overhead for the system, up to the point of being able to completely hang the whole desktop while the operation is running when it happens on the desktop folder.

An example is run this script inside the desktop folder: while true ; do touch testfile ; rm -f testfile ; done The desktop will freeze and won't be usable. It even won't be possible to stop the script itself from the graphical environment, you must change to a virtual text terminal and, there, kill the process.

This patch changes the way the events in the desktop are managed. Now, the first event launches the operation, as before. But if an event arrives while the operation is in process, it only sets a flag to indicate that the current operation is invalid and that it must be repeated. When the current operation ends, if the flag is not set the code knows that there were no changes during the operation, and can safely return; but if the flag is set, this operation is invalid, so it will wait 500 milliseconds (to avoid overloading the system) and will retry the operation. It is a must to put the delay after recreating the icons and not before, because recreating everything needs more than one second, and, since it is done by asynchronous functions, the time needed can have big variations.

This patch is a simplified alternative to !135 (closed) ; only one should be merged.

Fix part of #120 (closed)