Implement cart ROM writes as described by Jan Goldacker

[loganmc10]

Fixes #609

@jago85 did some investigation on how cart ROM writes are handled a while ago (#511 (comment))

The basic gist I got from that is that our current "rom_written" code should only happen if the game tries to read back the value when PI_STATUS_IO_BUSY is set, otherwise it should just read the regular value.

It does seem to work that way, this change fixes WCW Backstage Assault, and other games that depend on this code (A Bug's Life, Toy Story 2) still work.

A big thanks to @jago85, there is no way we would have figured this out without those actual hardware tests.

[loganmc10]

I'm not sure why AppVeyor isn't happy, but it doesn't seem related to this change. I've tested all the games that I know of that attempt to write to the cartridge ROM, I haven't found any regressions

[Jj0YzL5nvJ]

Yeah, AppVeyor has been having very weird issues lately.
All configurations are being compiled properly on my machine.

[loganmc10]

Hi all, any thoughts on this?

[Jj0YzL5nvJ]

See #729

[loganmc10]

Sorry I meant specifically related to this PR, I trust the AppVeyor issues will get sorted out

[bsmiles32]

Also tested said games and found no obvious regression.
I still think with some refactoring, we could move this IO_BUSY logic inside the PI instead of spreading it in cart/cart_rom. It would also probably best reflect how the hardware works. But for now it is better than what we have and fixes real issues. So merging this work.

Thanks to all involved especially to @jago85 for the hardware investigation.