Use Rectangle.ofFloat for bounds in WM_PAINT to avoid rounding issues

[arunjose696]

When resizing or expanding scrollable with styled text (e.g. Git revision information), the width of the newly painted region can be small such that they are non-zero in pixels but rounded down to zero in points (after pixelToPoint call). This causes StyledText::handlePaint to skip drawing due to the below check , resulting in visible artifact lines.

if (event.width == 0 || event.height == 0) return; link

Using float-based bounds preserves these non-zero regions from being rounded to zero in points and prevents missed paint operations during resize.

Steps to reproduce:

(Originally posted by @HeikoKlare in vi-eclipse/Eclipse-Platform#560)

• Open a current IDE on a 150% monitor
• Open any Java file under Git version control and show the revision information in the ruler (right click -> "Team" -> "Show revision information")
• Hover over any of the markers with changes, such that the window with detail information appears
• Slowing increase the width of that window
• Artifact lines like the one mentioned in Artifacts when resizing EGit blame window at 150% monitor zoom vi-eclipse/Eclipse-Platform#560 appears without this change,

[github-actions]

Test Results (win32)

   27 files   -  7     27 suites   - 7   5m 12s ⏱️ + 1m 2s
4 575 tests  - 57  4 504 ✅  - 55  70 💤  - 3  1 ❌ +1 
  110 runs   - 57    110 ✅  - 54   0 💤  - 3  0 ❌ ±0 

For more details on these failures, see this check.

Results for commit 92ec31a. ± Comparison against base commit ddac0aa.

This pull request removes 57 tests.

AllWin32Tests ImageWin32Tests ‑ testDisposeDrawnImageBeforeRequestingTargetForOtherZoom
AllWin32Tests ImageWin32Tests ‑ testDrawImageAtDifferentZooms(boolean)[1] true
AllWin32Tests ImageWin32Tests ‑ testDrawImageAtDifferentZooms(boolean)[2] false
AllWin32Tests ImageWin32Tests ‑ testImageDataForDifferentFractionalZoomsShouldBeDifferent
AllWin32Tests ImageWin32Tests ‑ testImageShouldHaveDimesionAsPerZoomLevel
AllWin32Tests ImageWin32Tests ‑ testRetrieveImageDataAtDifferentZooms(boolean)[1] true
AllWin32Tests ImageWin32Tests ‑ testRetrieveImageDataAtDifferentZooms(boolean)[2] false
AllWin32Tests ImageWin32Tests ‑ test_getImageData_fromCopiedImage
AllWin32Tests ImageWin32Tests ‑ test_getImageData_fromImageForImageDataFromImage
AllWin32Tests TestTreeColumn ‑ test_ColumnOrder
…

♻️ This comment has been updated with latest results.

[akoch-yatta]

LGTM, the change makes sense and fixes the issue as describes. The only risk I see is that the bounds could now be bigger (in pixels) as requested when scaling up again and could render too much instead of too less. With some testing I could not find any of those scenarios happening, as the OS is already correctly clipping the region to render correctly to e.g. 1 px