RFR(S) 8200109: NMT: diff_malloc_site assert(early->flags() == current->flags(), "Must be the same memory type")

[zgu at redhat.com]

There is possibility that all allocations from particular site are
freed, then have some allocations from the same site for another memory
type.

As indicated in bug report, this likely happens to some common utility
classes (e.g. GrowableArray), which have long calling chain leading to
allocation site. 
 

Bug: https://bugs.openjdk.java.net/browse/JDK-8200109
Webrev: http://cr.openjdk.java.net/~zgu/JDK-8200109/webrev.00/

Test:

  hotspot_nmt on Linux x64 (fastdebug and release)
  All passed except except
  runtime/NMT/CheckForProperDetailStackTrace.java (JDK-8218458) with 
  release build

Thanks,

-Zhengyu