Wrong 32 bit address mode used for 16 bit pointer

[Edilbert]

In the sequence:

lda [ab],z
sta (cd),z

the store operation falsely uses cd as a 32 bit pointer, instead of 16 bit as it should.

Putting another instruction in between can be used as workaround.

lda [ab],z
clv
sta (cd),z

[lydon42]

The fix does reset the signal to early, which disrupts the parallel memory process. I have a working fix build in matrixsid, for which I will now first write a complete four-way unittest (lda/sta, sta/lda, ldq/stq, stq/ldq).