[MIPS] Mark jalr[.hb] $zero, $ra as a return instruction

[CSharperMantle]

This PR marks MIPS jalr[.hb] $zero, $ra as a return instruction.

Resolves #7355.

Effects

Before:

After:

[CLAassistant]

All committers have signed the CLA.

[CSharperMantle]

Hi @psifertex and @zznop, I've rebased the branch onto the latest dev branch. Any progress on getting this reviewed? :)

[zznop]

Yes, I'll get this tested and merged today or tomorrow.

[zznop]

Looks like this code change affects jalr branches that don't use the zero operand too, and it causes regressions

Should be a simple fix. Looking into it now.

[zznop]

I'm going to go with:

		case MIPS_JALR_HB:
			if (instr.operands[0].reg == REG_ZERO && instr.operands[1].reg == REG_RA)
				result.AddBranch(FunctionReturn, 0, nullptr, hasBranchDelay);
			else
				result.delaySlots = 1;
			break;

I see that you were probably following the pattern for MIPS_JR/MIPS_JR_HB, but I think we want to continue to not do anything in cases where it's not a jalr $zero, $ra and let Binja figure out the branch type when lifting.