Closes #298

One of them is entirely spurious, but was easy to suppress by lightly
rearranging some code.  The other was a real NULL pointer deref, but
fortunately one that should never happen unless the xfdesktop-settings
binary is somehow corrupted.

This makes things consistent with our other projects, and will allow CI
to work properly to enable/disable them.

Stupid C; instead of an == test, I typo'd an = assignment.

This seemed to just never work reliably; most of the time when moving a
folder icon file into a subdir of the desktop, it wouldn't get picked
up.  And if the file was changed or deleted, that change wouldn't get
noticed.

263 translated messages.

Transifex (https://explore.transifex.com/xfce/).

It's still a little messy, but this cleans things up a little.  The
problem is that we could get into a situation where we're trying to
place items but we have no grid yet.  So now I'm simplifying things a
bit, and separating loading the items from the model from actually
placing them on the grid.  We also just protect actions that need a grid
with a check to see if we have a grid, since just because the widget is
realized, it doesn't mean it has a grid.

In addition, the previous code was using the widget's size request to
size the grid (and falling back to its preferred size otherwise).  This
is bonkers: we can only size the grid based on the widget's actual
allocation; anything else can give us sizes that are wrong.

Two issues:

* Any time a ViewItem was freed, the GtkRowReference instance (if any)
  associated with it was not freed.
* On finalizing the XfdesktopIconView, if a model was still attached,
  the ViewItem list would not get freed.

There were two issues:

1. Previously it would just prepend datadir, which isn't correct; both
   prefix and datadir need to be prepended.
2. If the option is specified as an absolute path, nothing should be
   prepended.

Closes #302.

This reverts commit 880579c3.

Now included in xfce/xfce4-dev-tools/ci/build_project.yml, see
xfce4-dev-tools!110.

262 translated messages, 1 untranslated message.

Transifex (https://explore.transifex.com/xfce/).

263 translated messages.

Transifex (https://explore.transifex.com/xfce/).

Closes #242.

Addresses #261.

Previously we'd invalidate the entire slot only if the item was under a
DnD operation.  Now we unconditionally invalidate the union of the slot
rect and the item's calculated extents, if any.

This might cause a *little* more drawing than necessary in some cases,
but fixes some intermittent drawing glitches when moving icons around.

Also chain up in size_allocate, since the parent impl does most of what
we were doing anyway.

The changes in a style update might affect where we draw the icons, but
shouldn't affect the pixel data in the icons themselves.

Closes #261.

There are far too many instances of gtk_widget_queue_draw() all over, I
suspect to paper over redraw issues.  Removing all of these (and
replacing them with individual or all icon invalidation) should be fine
to do.  There's still the possibility that some areas won't get
invalidated when they need to be, but they should be fixed when found.
The only time we should `_queue_draw()` is when the backdrop changes.

Closes #284.

Closes #260.

Closes #261.

Closes #291.

There was really no need for this, as the concrete impls were listening
to `notify::workarea`.  At the time, this just wasn't hooked up
properly.

Again, icon view should not be concerning itself with this; manager
needs to size and place the icon view in the correct position.

(Which we never do!)

All of these various changes should just be handled in one place; the
icon view itself should perhaps not have to worry about it, and
something external should place and size it.

This leaves screen-changed alone for now, but I'll revisit that soon.

Closes #289.

It currently isn't being used at all, and the icon view shouldn't be
concerning itself with that anyway.