I can't post the study due to copyright restrictions, but I think this falls under fair use (And I hope this works

)
Hey, not bad!
Note that CAPT has other studies that don't agree entirely with this, blah blah. There isn't a lot of research done with MBTI + IQ in the open field... so... YMMV.
For your specific questions;
Half of all people have IQ's in the top 50%, and half of these are N's, and half are S's. This is also true of the bottom half, except 3/4 of the S's in the lower half (so 3/8 of all S's total) are in the lowest 25%, and only 1/4 of N's in the lowest half are in the lowest 25% (so 1/8 of all N's total).
That would be roughly correct, although it wouldn't be 50/50 at IQ 100 (probably about 60/40 S/N). As above, about 85% of 89 IQ and lower are Ss, but it's an even split above 110.
If you were to take a sample of all the people in the top 50% for IQ, half would be N's and half S's. So 1/4 of the entire population in the top half are N's, and only 30% of the pop. are N's, then that would mean 5/6 of all N's are in the top half, and only 1/6 of all N's are in the lower half.
Top 25% in this case, half are Ns and Ss. Essentially, if this was normalised correctly, 12.5% of the population would be N and IQ>110... and 12.5% of the population would be S and IQ>110.
The sample is somewhat unevenly spread among IQ and N/S, so it's better to look at each as it's own set rather than relate it to the general population.