bug fix for diskpbb model

Ken Ebisawa supplied a fix to diskpbb.f. He notes that

  The bug was such that the old normalization was wrong in that it
 was 0.75/p times smaller.  The correct normalization is obtained
 by multiplying 0.75/p to the normalization of the old code. Disk
 radius is proportinal to the square root of the normalization, and p
 can be as low as 0.5, so the correct inner radius may be up to
 1.11 time greater than the radius obtained from the old code.
 Luckily, this would not dramatically change the consequences of the
 papers already published by old diskpbb code (Okajima et al.,
 Vierdayanti et al., and what else?).

Keywords: HEAsoft, xspec