minor performance tuning

#10

#9

• add canny (canny edge detector)
• specifiable erosion/dilation box size
• add libyuv comparing w/ cgo benchmark and imaging

#7

• add laplacian
• add gradient
• add highpass
• add erosion
• add dilation
• improve grayscale performance

#6

• add laplacian filter
• update JIT benchmark results to be easier to read.

#5

• add Rotate

#4

• add emboss
• adapt CPU feature.

#3

• tune integer divide
• alloc/free internal api change
• add invert func

#2

#1