patch for previously reported polsturm bug

["Carl R. Witty" <cwitty@newtonlabs.com>]

Here's a patch to fix the bug I just reported against polsturm.

After the patch, I get the correct answer for the previously buggy
case:

(13:28) gp > polsturm(x, 0, 1)
%1 = 0

--- polarit2.c~ Mon Feb  7 07:29:08 2000
+++ polarit2.c  Sun May 14 13:28:12 2000
@@ -2456,6 +2456,7 @@
   {
     s = b? gsigne(poleval(x,b)):  sl;
     t = a? gsigne(poleval(x,a)): -sl;
+    if (t == 0) return 0;
     avma = av; return (s == t)? 0: 1;
   }
   u=gdiv(x,content(x)); v=derivpol(x);