[PartitionAlloc] Don't poison memory on free.

It might? be the cause of a perf regression.

Bug: 882471,680657
Change-Id: Iec34a348426ccfdf5864a3b2cbe0c377da20ac0e
Reviewed-on: https://chromium-review.googlesource.com/1220468
Commit-Queue: Chris Palmer <palmer@chromium.org>
Reviewed-by: Emil A Eklund <eae@chromium.org>
Reviewed-by: Kentaro Hara <haraken@chromium.org>
Cr-Commit-Position: refs/heads/master@{#590508}

base/allocator/partition_allocator/partition_page.h

@@ -209,13 +209,9 @@ ALWAYS_INLINE void PartitionPage::Free(void* ptr) {
   PartitionCookieCheckValue(ptr);
   PartitionCookieCheckValue(reinterpret_cast<char*>(ptr) + slot_size -
                             kCookieSize);
-#endif
 
-  // Perhaps surprisingly, this does not measurably regress performance. See
-  // https://crbug.com/680657 for history. We formerly did this in
-  // |PartitionFree|, and in `DCHECK_IS_ON()` builds we redundantly did it
-  // here, too. Now we only do it here, unconditionally.
   memset(ptr, kFreedByte, slot_size);
+#endif
 
   DCHECK(this->num_allocated_slots);
   // TODO(palmer): See if we can afford to make this a CHECK.