Argb support fix

[Drvi]

Fixes #8

In the case of BGR, ARGB, BGRA and ABGR represented with floating points we were handling the struct pixel layout twice -- once with libng by calling png_set_brg and png_set_swap_alpha and by converting the data layout to RGBA in _prepare_buffer -- these two operations sort of cancelled each other out. The fix implemented here is not to do the conversions.

cc: @johnnychen94

[codecov]

Codecov Report

Merging #9 into master will increase coverage by 1.56%.
The diff coverage is n/a.

@@            Coverage Diff            @@
##           master      #9      +/-   ##
=========================================
+ Coverage   64.84%   66.4%   +1.56%     
=========================================
  Files           4       4              
  Lines         256     256              
=========================================
+ Hits          166     170       +4     
+ Misses         90      86       -4

Impacted Files	Coverage Δ
src/io.jl	91.41% <ø> (+2.45%)	⬆️

---

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 69d5cbc...465cd86. Read the comment docs.

[IanButterworth]

I tried to figure out the resolver error but I'm failing.
Any ideas @johnnychen94?

[johnnychen94]

Thanks for the lighting fix!

A new ColorVectorSpace release with compat fix is needed, see also JuliaGraphics/ColorVectorSpace.jl#122 and JuliaImages/ImageCore.jl#124

It succeeded in nightly because of a Pkg issue, possibly related to JuliaLang/Pkg.jl#1720

This is only a test stage issue, and if it's tested locally with a customed version of ColorVectorSpace, I think it's okay to not wait for a new version of ColorVectorSpace

[kimikage]

BTW, PNGFiles.jl uses a primitive Array as the buffer, so I doubt that we really need to use colorview or depend on ImageCore.jl.

Of course, it is intuitive that PNGFiles.jl depends on ImageCore.jl. However, we faced the dependency issue, so I think the low-level processing looks reasonable.

Edit:
Of course, my proposal above is not a solution of this issue.

[IanButterworth]

I removed the Images dep during test by absorbing absmaxfinite in #10 and rebased this