Fix reset_transform in Cocoa/iOS

[HalfWhitt]

Fixes #4044

Adding the test first...

So far I have a fix that seems to work for Cocoa, but breaks iOS in new and interesting ways.

PR Checklist:

• All new features have been tested
• All new features have been documented
• I have read the CONTRIBUTING.md file
• I will abide by the code of conduct

[HalfWhitt]

This is odd... from both Linux Wayland Gtk4 tests (with and without adwaita):

 File "/home/runner/work/toga/toga/testbed/build/testbed/ubuntu/noble/testbed-0.0.1/usr/lib/testbed/app/tests/widgets/test_canvas.py", line 715, in test_reset_transform
    screenshot = canvas.window.as_image(format=Image.Image).crop((0, 0, 200, 200))
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/toga/toga/testbed/build/testbed/ubuntu/noble/testbed-0.0.1/usr/lib/testbed/app_packages/toga/window.py", line 619, in as_image
    return Image(self._impl.get_image_data()).as_format(format)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/toga/toga/testbed/build/testbed/ubuntu/noble/testbed-0.0.1/usr/lib/testbed/app_packages/toga/images.py", line 194, in __init__
    raise TypeError("Unsupported source type for Image")

And the screenshot on Linux Wayland Gtk3 is just... entirely transparent. Great.

Edit: didn't realize some platforms don't support Window.as_image(). That's fine, I should only be testing that step on Cocoa and iOS anyway.

[corranwebster]

This is also true for the identity transform and a few others which are their own inverses (like (-1, 0, 0, -1, 0, 0) and (-1, 0, 0, 1, 0, 0)).

Edit: to be constructive, I think it suffices to check if the determinant a * d - b * c == 0.

[HalfWhitt]

Hm. As I understand it, if the transform (relative to the original) is the identity matrix, then the current would equal the original and we would've already exited. But I hadn't actually realized other matrices are their own inverse as well, thanks! Math is wild.