Use DMA engine to blit to framebuffer

[blitz]

The ULX3S has a DMA engine to offload copying pixel data to the framebuffer. This should result in smoother display and less CPU usage.

                dma@80000 {
                        compatible = "spinal,lib-dmasg";
                        reg = <0x80000 0x10000>;
                        interrupt-parent = <0x03>;
                        interrupts = <0x0c>;
                        #dma-cells = <0x01>;
                        #address-cells = <0x01>;
                        #size-cells = <0x00>;
                        phandle = <0x04>;

                        dma-channel@0 {
                                reg = <0x00>;
                                interrupt-parent = <0x03>;
                                interrupts = <0x0c>;
                        };

                        dma-channel@1 {
                                reg = <0x01>;
                                interrupt-parent = <0x03>;
                                interrupts = <0x0d>;
                        };
                };

Depends on #6.