The MBC5 isn't really all that complicated.
This reference along with
this C++ implementation should be more than sufficient to pull it off. xzakox apparently got his working after I started talking about his
old abandoned thread and then he showed up here out of nowhere, but after reporting success, he fell off the face of the earth again and never released his complete VHDL code. This next school year, I plan to finally get around to getting my cart design manufactured so I can write my own Verilog, but that will probably have to wait until I find a new job, since I just lost the one I had, and PCB's aren't exactly in the budget right now.
To help get you started, I've looked into the different chips available, and I came to the same conclusion that xzakox did in terms of part choice, specifically the Xilinx XC9536. It's an obsolete part, but the replacement XC9536XL is 5v tolerant (though it's a 3.3v part). So, if you're planning to go into mass-production, you probably want to go 3.3v anyway, because it's pretty nearly impossible to find 5v parts anymore in the sizes you'll want. So, if you go 3.3v, Micron makes good Flash ROMs in 32/64Mbit sizes, and you can get Fe-RAM from Cypress up to 1Mbit with an 8-bit data bus (up to 4Mbit with a 16-bit data-bus, but you'd have to multiplex the data bus), then you wouldn't need a battery.