Optimize bokeh blur convolution

[Sergio0694]

Prerequisites

• I have written a descriptive pull-request title
• I have verified that there are no overlapping pull-requests open
• I have verified that I am following matches the existing coding patterns and practice as demonstrated in the repository. These follow strict Stylecop rules 👮.
• I have provided test coverage for my change (where applicable)

Description

Optimized the bokeh blur effect, with a more specialized variation of the changes introduced in #1465.

Benchmark

EDIT: updated benchmark after 2e53a44.

EDIT: updated benchmark after e4ba017.

EDIT: updated benchmark after 442e467 and c38fc81.

[Sergio0694]

Investigating the test failures with the effect on smaller image regions, and also noticed a previously existing bug that was likely causing some out of bounds writes in the past, very weird we hadn't spotted that one before. Glad it popped up now! 😄

EDIT: fixed in 7db1225 🎉

[Sergio0694]

Removed the WorksWithDiscoBuffers test for the bokeh blur processor, as it was just always throwing due to insufficient memory. @JimBobSquarePants I guess you could reintroduce it later on if you wanted to, but for now I wanted to have the CI run to verify all the other tests and the changes in this PR. Also, noticed that eg. the gaussian blur doesn't have this test either, so... 🤷‍♂️

ImageSharp/tests/ImageSharp.Tests/Processing/Processors/Convolution/BokehBlurTest.cs

Lines 177 to 183 in ff94d20

	[Theory]
	[WithTestPatternImages(100, 300, PixelTypes.Bgr24)]
	public void WorksWithDiscoBuffers<TPixel>(TestImageProvider<TPixel> provider)
	where TPixel : unmanaged, IPixel<TPixel>
	{
	provider.RunBufferCapacityLimitProcessorTest(41, c => c.BokehBlur());
	}

[JimBobSquarePants]

Removed the WorksWithDiscoBuffers test for the bokeh blur processor, as it was just always throwing due to insufficient memory.

This is concerning. We weren't getting any throws for this in previous runs.

[codecov]

Codecov Report

Merging #1475 (a8cae3f) into master (ff94d20) will increase coverage by 0.06%.
The diff coverage is 97.10%.

@@            Coverage Diff             @@
##           master    #1475      +/-   ##
==========================================
+ Coverage   83.49%   83.55%   +0.06%     
==========================================
  Files         742      741       -1     
  Lines       32347    32462     +115     
  Branches     3639     3648       +9     
==========================================
+ Hits        27009    27125     +116     
  Misses       4625     4625              
+ Partials      713      712       -1     

Flag	Coverage Δ
unittests	83.55% <97.10%> (+0.06%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
...s/Convolution/Convolution2PassProcessor{TPixel}.cs	100.00% <ø> (ø)
...essing/Processors/Convolution/KernelSamplingMap.cs	94.11% <33.33%> (-5.89%)	⬇️
src/ImageSharp/Common/Helpers/Numerics.cs	97.65% <95.31%> (-1.01%)	⬇️
...ssing/Processors/Convolution/BokehBlurProcessor.cs	100.00% <100.00%> (ø)
...ocessors/Convolution/BokehBlurProcessor{TPixel}.cs	99.34% <100.00%> (+0.49%)	⬆️
...mageSharp/Common/Helpers/SimdUtils.HwIntrinsics.cs	99.56% <0.00%> (+1.30%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ff94d20...a8cae3f. Read the comment docs.

[Sergio0694]

Reverted to draft as I just had some new ideas for some other possibly juicy optimizations to add 😋

[Sergio0694]

Alright, happy with the progress so far, I'd say this PR is good for review, and we can always revisit later on 🎉

[JimBobSquarePants]

LGTM. Very nice improvement! 🚀