There's no practical utility in this unless I want to make the palette size variable,
and this makes performance worse updating the screen when the palette changes,
but I'm on a mission to remove `from textual_paint.paint import PaintApp`
since it can't be at top level due to cyclic imports, and it has
side effects when importing `textual_paint.args` in turn.
My TODO comment said "add an attribute to ToolPreviewUpdate or make it's x/y Optional", which is both bad grammar ("it's" vs "its") and a bad idea, since the Canvas isn't involved in this update, so its event definition shouldn't have to be complicated by it.
I like this much better, just splitting the event handler into two functions, giving room to express the optional nature of the coordinate in the signature, and avoiding constructing two message objects at both of the callsites.
If git blame tools were smarter/easier to use, or if I was using a node-based programming paradigm, I would've never made this mistake. I would have done the natural and correct thing in the first place, but instead, I tried to avoid indenting a large block of code, which generates a noisy commit, and a barrier when using git blame.
There are many more failures now, which I should probably take a look at
at some point, but I don't know how to best inspect the changes yet.
I'll mark these as xfail (expected to fail) for now.
...but NOT the Free-Form Select test! I'm noticing that when marked
as skipped, running with --snapshot-update deletes the snapshot.
So neither xfail or skip (or xfail with run=False) is a good solution.
It's almost as if people don't normally do TDD with snapshots.
Anyways, HOPEFULLY I can move on now, to actually fixing things,
like the Polygon tool, and new bugs I've found while adding tests.
