We sort of gave up back in bug #11821, but it turns out we can
do somewhat better, if the "%0" is replaced with "%I0%0" for
each SDT_V2 operand slot, for powerpc only.  This would cause
literals to be prefixed with "i", resolving the ambiguity.

Then in turn the translator's $nnnn regex matcher (tapsets.cxx:4675ish)
could accept "i" as a literal number prefix, not just "$".

Up to this point, it's all compatible with SDT_V2 binaries & logic.

However, the later REGISTER case can't presume that the lack-of-"i"
means a register number, since SDT_V2 binaries may exist that precede
this particular change.  So a later magic number may need to be assigned
to communicate to the translator that %I0%0 was specified, and thus
lack-of-"i" means "register number, for sure".