starange "limit" behaviour



limit does handle a^x/b^x correctly (with questions), so the 7^x/8^x
case is not a deep problem, but some sort of coding error.

Similarly, 

> taylor(7^(n*n)/8^n, n, inf, 1) =>  - LOG(8) n + 1 + . . .

appears to be a shallow bug, since taylor(x^2+x+1,x,inf,1) gives all
three terms.  And taylor(7^(n^2)...,n,inf,2) gives them as well.

But even if they're shallow, someone has to find the problem in the code....

          -s