Support for CUDA (New Subpackage)

[ChantalJuntao]

While working on improving the speed of my image registration analysis, I've noticed that ImageTransformations doesn't currently support compatibility for GPU computing using CuArrays. Therefore, I've made a first attempt at a GPU version of ImageTransformations.warp that supports CuArrays that can be wrapped in Offset Arrays and defaults to linear b-spline interpolation.

I'm still working on trying to get this to use the actual Interpolations package to support more functionality, and my test suite is still quite incomplete. I'm also not sure if this will work on grayscale or colored images yet. Any help or feedback would be extremely appreciated.

[codecov[bot]]

Codecov Report

Merging #156 (c565dfe) into master (631d63b) will decrease coverage by 0.18%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #156      +/-   ##
==========================================
- Coverage   90.00%   89.81%   -0.19%     
==========================================
  Files           8        8              
  Lines         220      216       -4     
==========================================
- Hits          198      194       -4     
  Misses         22       22              

Impacted Files	Coverage Δ
src/autorange.jl	93.18% <0.00%> (-0.16%)	:arrow_down:
src/warp.jl	96.29% <0.00%> (-0.14%)	:arrow_down:
src/resizing.jl	100.00% <0.00%> (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 631d63b...c565dfe. Read the comment docs.

[timholy]

FWIW we agree this needs additional work but is not high priority, as it turns out that warping is not a real bottleneck in the analysis we were planning to optimize. To be continued later. In the meantime, let's leave this open in case it inspires someone else.