TODO: * Multiprocess capabilty using either: - threads - multiple instances using shared memory to share a cache - Asynchronous via asio or libevent * ICC profile integration via lcms library * Lossless Rotation / transposition support for JPEG tiles * JPEG source image support * Look into using malloc_usable_size to trace real allocated space * Lanczos, bilinear etc interpolation for CVT * Copy EXIF, IPTC data for CVT exports * Optimize FIF code for very large files * Rewrite JPEG writer code for better buffered output