render: Handle negative dest point values in apply_filter properly

[SuchAFuriousDeath]

Fixes #22792 (and maybe a handful of other issues).

The image is taken from Ruffle itself, hence the no outliers. If taken from Flash, there are like 22k outliers (But visually it looks the same, probably just a filter implementation thing). Not sure which one to pick here.

[torokati44]

If taken from Flash, there are like 22k outliers

At what tolerance tho?
🙄 #22008

[SuchAFuriousDeath]

Dunno. I guess I should try a higher tolerance? I haven't looked into these things too much. The fix should be correct though.

[kjarosh]

The image is taken from Ruffle itself, hence the no outliers. If taken from Flash, there are like 22k outliers (But visually it looks the same, probably just a filter implementation thing). Not sure which one to pick here.

Always pick the one from FP; if Ruffle is not exact, add tolerance, outliers, multiple checks, etc. This prevents future PRs from changing the image all over the place and makes us exact in comparing FP with Ruffle (and not doing an ocular analysis only).

[SuchAFuriousDeath]

Lmao, you broke it 😂 Give me a while, I'll fix it.

[SuchAFuriousDeath]

Here is a full list of issues that need to be retested btw. https://github.com/ruffle-rs/ruffle/issues?q=is%3Aissue%20state%3Aopen%20%22would%20end%20up%20overrunning%20the%20bounds%20of%20the%20Destination%20texture%20of%22