Fix for ellipse_im with nonzero rotation

[cjblocker]

Test code:

import MIRT: ellipse_im, jim
import ImageTransformations: imrotate, Linear
jim([ ellipse_im(256,256; rot=0); 
      ellipse_im(256,256; rot=45);
      imrotate(ellipse_im(256,256; rot=0), π/4, (256,256), Linear(), 0.0)], colorbar=:none)

Before on the current HEAD of master:

After changes:

[johnnychen94]

Out of curiosity, I did a small benchmark on this and TestImages. It's not equivalent but at least it indicates some room for optimization.

julia> @btime ellipse_im(256,256; rot=0);
  11.531 ms (714 allocations: 101.42 MiB)

julia> @btime TestImages.shepp_logan(256);
  1.103 ms (4 allocations: 516.33 KiB)

[codecov]

Codecov Report

Merging #106 (bb280cc) into master (b66be11) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #106   +/-   ##
=======================================
  Coverage   99.81%   99.81%           
=======================================
  Files          44       44           
  Lines        2181     2182    +1     
=======================================
+ Hits         2177     2178    +1     
  Misses          4        4           

Impacted Files	Coverage Δ
src/fbp/ellipse_im.jl	100.00% <100.00%> (ø)
src/regularize/diffl.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 b66be11...bb280cc. Read the comment docs.

[JeffFessler]

@cjblocker turns out this was a bug in Matlab MIRT too so you've provided a double fix. thanks!
JeffFessler/mirt#17

@johnnychen94 stay tuned for a more sophisticated version in JuliaImageRecon...