[Split out from JDK-8217291]

In a few places of hotspot a possible failure of realloc isn't handled correctly: Either the return value isn't checked for NULL, of the original memory isn't freed, or both.

jdk.hotspot.agent/windows/native/libsaproc/sawindbg.cpp: m_msgBuffer = (char*) realloc(m_msgBuffer, len + strlen(m_msgBuffer));