Tiling optimizations

• Tile in cava to vectorize gamut fused channel loop
• Fix tiling in BFS to use correct loop order
• Tile further in BFS for vectorization
• Put init loop on GPU for BFS

requested review from @aaronjc4

assigned to @rarbore2

enabled an automatic merge when all merge checks for 1c3bc6d5 pass

mentioned in issue #29 (closed)

mentioned in commit 8e7b89c5

merged