av1_rc_update_rate_correction_factors: avoid log(0)

avoids a floating point error being generated. this doesn't change
behavior, the -HUGE_VAL return was being ignored due to the min()

BUG=aomedia:388

Change-Id: I698aab248c777b02b60081bb7619e0aa6f3a0ba2
diff --git a/av1/encoder/ratectrl.c b/av1/encoder/ratectrl.c
index 2141f30..56eae7e 100644
--- a/av1/encoder/ratectrl.c
+++ b/av1/encoder/ratectrl.c
@@ -444,8 +444,12 @@
 
   // More heavily damped adjustment used if we have been oscillating either side
   // of target.
-  adjustment_limit =
-      0.25 + 0.5 * AOMMIN(1, fabs(log10(0.01 * correction_factor)));
+  if (correction_factor > 0) {
+    adjustment_limit =
+        0.25 + 0.5 * AOMMIN(1, fabs(log10(0.01 * correction_factor)));
+  } else {
+    adjustment_limit = 0.75;
+  }
 
   cpi->rc.q_2_frame = cpi->rc.q_1_frame;
   cpi->rc.q_1_frame = cm->base_qindex;