1) The outer wheel,not the inner one ratchets with freewheels
2) A single CVT cannot play diff. no matter what it is
3) Dual belt type CVT's would seem problematic if they were speed controlled,if torque&speed controlled it should be a different matter,I think the argument that the overall efficiency is the product of the individual ones is incorrect (each 0.9 would get you 0.81 overall).It would be somewhat lower but not much,but unless they were identical, one would be doing the heavy lifting unless they were torque sensitive,It's a can of worms,stick with freewheels.
4) I assume your engine has an internal reduction.Assuming 20" wheels the speed is
S=0.058x rpm/R in mph,R is total reduction from crankshaft ,with an internal reduction of 5? , R becomes 5x38/18= 10.55 and at 5000rpm we get S=27.5 mph,assuming a max torque of 2.5 Nm
(1.8 ft.lbs ).The torque at the wheels would be 1.8x10.55 xeff (0.85?) =16ft.lbs, or 20lbs tractive force.An angle of 12 degrees is extremely steep,corresponds to a 21% grade.You must have got your numbers wrong.Assuming 450lbs total weight,you could climb a 20/450=0.044 grade or 2.5degrees,With a CVT and max ratio of 3 say (3-1.5) reduction you could do 7.6 degrees.Your speed at 5000rpm would be 27.5/1.5=18.3 mph.
All this is predicated of course on the 1.8 ft.lbs torque assumption (1.74 HP at 5000 rpm)