Copy screen + colour RAM to the left one place

This sequence of instructions:

loop:
         INY
         LDA ($D1),Y
         DEY
         STA ($D1),Y
         INY
         LDA ($F3),Y
         DEY
         STA ($F3),Y
         INY
         CPY $D5
         BNE loop

is the obvious any minimal way to copy the screen and colour RAM data.
The constants $D1 and $F3 are fixed by the publicly documented interface
of the KERNAL to point to the screen and colour RAM of the screen.

The Y register is the only register that can be used to post-increment an
indirect address on the 6502.

Thus the only efficient manner to copy a string to the left one byte
is to read, decrement Y by one to get the destination address, store,
then increment Y again so that it points to the source offset, now read
the colour RAM byte, add one to Y so that we point to the destination,
and then write it.  

Similarly the check against $D5 (current length of line) is also
necessary for working out when to terminate the loop.  

In fact, about the only variant possible would be to swap the $D1 and $F3
constants so that the colour RAM is copied before the screen RAM in each
loop, instead of after it.

It is telling that this routine was the result of independent implementation
of this routine, without looking at the KERNALs routine, in terms of supporting
the argument that the routine is the obvious manner in which to do this.