Change veryLargeRect size to fix issue with masks (#1880)

[nevil]

When using scaleToFill / scaleAspectFit / scaleAspectFill mode with an animation that has a subtract mask, the animation could sometimes fail to render. veryLargeRect is used as the size of the "bounds" rect from which the path is subtracted.

Alternative considered
I digged through the old code and found that the old Objective-C implementation actually passed in the bounds size.
I guess that could be an alternative solution if needed.

- (void)updateForFrame:(NSNumber *)frame withViewBounds:(CGRect)viewBounds {
   if ([self hasUpdateForFrame:frame]) {
     LOTBezierPath *path = [_pathInterpolator pathForFrame:frame cacheLengths:NO];

     if (self.maskNode.maskMode == LOTMaskModeSubtract) {
       CGMutablePathRef pathRef = CGPathCreateMutable();
       CGPathAddRect(pathRef, NULL, viewBounds);
       CGPathAddPath(pathRef, NULL, path.CGPath);
       self.path = pathRef;
       self.fillRule = @"even-odd";
     } else {
       self.path = path.CGPath;
     }

     self.opacity = [_opacityInterpolator floatValueForFrame:frame];
   }
 }

502b18ff#diff-233d9c9fcd51dcb99d9884451e850073e2712d14d3c05e0e91bae1c551abe91eR42

Video capture before/after the change

iPhone14 Pro Simulator

Before	After
Simulator.Screen.Recording.-.iPhone.14.Pro.-.2022-12-29.at.15.41.19.mp4	Simulator.Screen.Recording.-.iPhone.14.Pro.-.2022-12-29.at.15.41.47.mp4

[nevil]

How large is a large rect? 🤷‍♂️

[calda]

Thanks for the fix! Nice find.

[nevil]

Seems like the same test is failing each time, but it shouldn't really be affected by the change to veryLargeRect as it isn't used in that test. 🤔

✖ testParsing_simpleAnimation, XCTAssertEqualWithAccuracy failed: ("1.3376318674468077") is not equal to ("2.0") +/- ("0.65")

[calda]

The performance tests are flaky sometimes -- we just need to increase the threshold