Sure DDR is "better", but that's not a contest. That's about selecting an appropriate solution for the requirements.
Of course it is - it's about finding the best solution to a problem. I'm yet to see a hardware platform which would suffer from having too much memory capacity or speed, so the more the merrier.
No, that's not gonna require 80 IOs, but 50-odd IOs, actually.
It's more like 60, but also remember that unlike in DDRx, the entire bus has to be length-matched - all signals (!!!). In DDR address/control works independently of data lanes, and each data byte lane can be length matched independendly of others, so it's actually easier to route (even if we ignore the fact that DDRx requires only half as much data width to achieve the same bandwidth).
DDR controllers are definitely harder to design, the MachXO2 is a very humble FPGA, and Lattice does not offer memory controllers for free.
Except for DDR3 with it's peculiarities like link training, I would say the rest are more-or-less the same, provided that FPGA comes with the hardware to make PHY possible, and as far as my (admittedly very rudimentary) research has shown, Lattice does provide PHY module generators for free, so it's only the controller part which needs to be designed. Not easy for sure, but not that hard either. In my experience designing a LPDDR1 memory controller, the hardest part (after PHY that is) was higher level functions like request reordering, command scheduling and other functions one would need for a full-featured AXI slave port with bursts support. Video applications typically have a very straightforward access pattern, and so they usually don't need sophisticated bus interface adapters.