double precision calculation using CUDA library

• max widget size setted to same as FHD (1080p, 1920 x 1080)
• device memory & pinned host memory allocated with above fixed max size to avoid reallocations
• much slower than float openGL fragment shader version when high iteration pixels are visible
• more cuda core -> faster texture generation
• openGL's PBO(PixelBuffer Object, data async transfer between host-device) may be appliable for optimization

• mouse left button 2-dimensional drag : change complex constant
• alt + mouse left button vertical drag : change real exponent
• mouse right click : toggle log_expression