Much Faster sRGB Companding

[JimBobSquarePants]

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

Updates the bulk sRGB Companding method to use the similar LUT approach as MagicScaler (thanks @saucecontrol ! )

Results are awesome. Massive speedup in real world scenarios.

I noticed some oddities (I'd previously introduced) in the existing unit tests during the work which I cleaned up also.

[antonfirsov]

@saucecontrol can we use this approach for pixel blending? Seems cheaper than #1433.

[codecov]

Codecov Report

Merging #1481 (5659148) into master (611ff85) will increase coverage by 0.03%.
The diff coverage is 97.16%.

@@            Coverage Diff             @@
##           master    #1481      +/-   ##
==========================================
+ Coverage   83.48%   83.52%   +0.03%     
==========================================
  Files         741      741              
  Lines       32575    32664      +89     
  Branches     3652     3661       +9     
==========================================
+ Hits        27194    27281      +87     
- Misses       4668     4669       +1     
- Partials      713      714       +1     

Flag	Coverage Δ
unittests	83.52% <97.16%> (+0.03%)	⬆️

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

Impacted Files	Coverage Δ
...ImageSharp/PixelFormats/Utils/Vector4Converters.cs	100.00% <ø> (ø)
...ssing/Processors/Transforms/Resize/ResizeWorker.cs	98.73% <50.00%> (-0.02%)	⬇️
src/ImageSharp/Common/Helpers/Numerics.cs	96.99% <60.00%> (-0.82%)	⬇️
...mageSharp/ColorSpaces/Companding/SRgbCompanding.cs	100.00% <100.00%> (ø)
...ssing/Processors/Transforms/Resize/ResizeKernel.cs	100.00% <100.00%> (ø)

---

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 611ff85...5659148. Read the comment docs.

[antonfirsov]

Let's make the lazy code thread-safe + a major concern regarding tests.

Can't comment on the SIMD code (too much time to review properly), but I trust the test coverage as usual.

[saucecontrol]

@saucecontrol can we use this approach for pixel blending? Seems cheaper than #1433.

Yep, I use it for all my linear pixel conversions. It also works great in combination with fixed-point math, which I still use as a fallback on non-RyuJIT netfx. In that case, you can have a 15-bit LUT with direct lookups and no interpolation.