Fix disassembler tracing early exits with some LD instructions

[parasyte]

This took a long time to track down. I'm not entirely certain what the real solution should be, but this seems to fix it.

These screenshots show the problem. The disassembly prematurely ends at an LD instruction:

I found the problem was that None was being returned as the first Cursor for this specific instruction pattern.

While tracking this down, I also noticed a bizarre use of RefCell that isn't needed. So I removed that in a second commit.

[Rodrigodd]

Thanks! I believe that is the correct solution.

The code is trying to detect if a bank switch happens at that instruction, but the value is unknown so is uncertain if that will happen or not. But other parts of the code already makes the assumption that a program will very unlikely change the bank under itself, so this fix is fine.