Investigate how to make the CPC port faster

[TheFakeMontyOnTheRun]

It's possible we could cache the line starts or create a LUT for each line in the 3D framebuffer.
Not only that, but given the memory we have, multiple pages could be used. It was tested with the double buffer allocated (but not used) and it worked fine

[TheFakeMontyOnTheRun]

Fixed with #91