Reduce XY Table to only using fixed, single template map, instead of fresh calculating one the size of the matrix every begin()

[connornishijima]

Keep a template map for a single display's worth of XY lookup table in PROGMEM, and do extra calculations during xy() calls to loop over it and discover the 1D indices of 2D coordinates in displays of any size, versus the current method of keeping an entire int16_t[matrix_width * matrix_height] map in RAM.