Fix sigfigs(n) rounding to work when the input is bigger tnan 10**n.

For example rounding 1234567/1 to 6 significant figures would fail.

libgnucash/engine/gnc-rational.cpp

@@ -171,7 +171,9 @@ GncRational::sigfigs_denom(unsigned figs) const noexcept
         ++digits;
         val /= 10;
     }
-    return not_frac ? powten(figs - digits - 1) : powten(figs + digits);
+    return not_frac ? 
+            powten(digits < figs ? figs - digits - 1 : 0) : 
+            powten(figs + digits);
 }
 
 GncRational