Improve DLDI on ARM7

[AntonioND]

Filesystem performance when DLDI runs on the ARM7 is bad. For starters, the code is in main RAM. We could move it to WRAM when it is used from the ARM7 so that both CPUs don't access main RAM at the same time.

This issue is here to keep track of the required changes to improve its performance.

@asiekierka: "If you're going to stick to my code, remember to then allow the sector cache to use all of the ARM9's DLDI area in main RAM."

I guess we can use malloc() on the ARM7 and move DLDI there if the user requests it? Or we could provide an alternative linkerscript that places it in WRAM directly...

This will need some careful thought.