PPC pooled data references: Add support for subi, addis, and subis instructions

[LagoLunatic]

Fixes a bug where the pooled reference calculation would mistake the subi instruction for addi because they have the same opcode and are only distinguished by their mnemonic string.

Before:

After:

Also adds support for the addis and subis instructions, which can sometimes be used by the compiler for very large data pools:

[encounter]

If the code operated on the basic form of the instruction instead of simplified, this wouldn’t be an issue. But LGTM for a quick fix

[LagoLunatic]

If the code operated on the basic form of the instruction instead of simplified, this wouldn’t be an issue. But LGTM for a quick fix

That's a good point.

I also just checked the flow_analysis code and it looks like there's a similar pattern there that I think could theoretically cause a bug for stringBase, but I can't find any code that uses subi on a stringBase so it may not be an issue in practice.

Might be worth refactoring both of them to not use simplified form anymore.