tay+tya is 4 cycles/2 bytes. lda (dp),y is 7 cycles/2 bytes
only ora #$20 IF the target character is lowercase. order the comparisons so accumulator can be successively modified. worst case, 2 loads are needed per word "##" -> load, no ora "A#" -> ora $0020 "#A" -> reload, ora #$2000 "AA" -> ora $#2020