Port of AGG 2.4

• Rewamp a bit the APIs names in C for consistency.
• Add additional compositions mode to match gd 2.4 modes
• SIMD implementation for compositions or other ops where it actually brings improvements. More to do there later especially for floating point bufs bgra128 where it is not yet the best implementation details to fully use avx2 or arm.
• this repo will stay standalone and used by a libgd branch to replace ftraster based rasterizing and will target gd 3 or 2.5+