It seems within reason to do, how much work it requires is a different matter. I saw someone compiled it for Switch earlier but performance was not great which means even more fun for the 3ds.
The PC stuff means you can probably figure out where to change to sort the controls, saves, and graphics location issues. The trouble will probably come with Mario 64 doing quite a bit with floating point numbers, which the 3ds can do more with than the likes of the DS (see many of the issues getting various quake games to run there) but you still don't want to be operating with them as much as Mario 64 does (watching all those videos on A press runs you can see how much it does). That would require some fairly decent tweaks right at the heart of the game, though all still within reason and I don't know if there would be a suitable less accurate mode to run in for the 3ds that might do better.
The PC port stuff also looks like it uses brute force to get around some of the hardware specifics as well (any game maker that approaches things like I saw in that code for the PC since about 30 years ago would probably have been fired on the spot, but would be perfect valid in the console world) so you would probably want to unpick that and do something better too.
I don't know if anybody will bother -- the 3ds had a pretty weak homebrew scene and it is pretty much dead at this point. If someone wanted to though then I could see it happen in a reasonable timeline (mario 64 DS was a commercial effort with probably many devs all funded to do it, homebrew devs tend to be doing it for fun in their free time after all). On the other hand people do really seem to love this game (personally I think most of Rare's stuff was better but I am in something of a minority on this one) and the mere fact we have a decompilation in space year 2020/2019 (rather than space year 2024 or so which was when I was expecting decompilation to get more practical going by past developments in the field) speaks to something.