MFV: libb2: use native calls for secure memory clearance

Drop our local patch and restore full vanilla upstream code in
contrib/libb2.

No functional change intended.  explicit_bzero() should continue to be used.

Obtained from:	libb2 b4b241a34824b51956a7866606329a065d397525
Sponsored by:	Dell EMC Isilon

sys/contrib/libb2/blake2-impl.h

@@ -131,11 +131,20 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c )
 /* prevents compiler optimizing out memset() */
 static inline void secure_zero_memory(void *v, size_t n)
 {
-#ifdef __FreeBSD__
-  explicit_bzero(v, n);
+#if defined(_WIN32) || defined(WIN32)
+  SecureZeroMemory(v, n);
 #else
-  static void *(*const volatile memset_v)(void *, int, size_t) = &memset;
-  memset_v(v, 0, n);
+// prioritize first the general C11 call
+#if defined(HAVE_MEMSET_S)
+  memset_s(v, n, 0, n);
+#elif defined(HAVE_EXPLICIT_BZERO)
+  explicit_bzero(v, n);
+#elif defined(HAVE_EXPLICIT_MEMSET)
+  explicit_memset(v, 0, n);
+#else
+  memset(v, 0, n);
+  __asm__ __volatile__("" :: "r"(v) : "memory");
+#endif
 #endif
 }
 

sys/crypto/blake2/config.h

@@ -17,3 +17,5 @@
 #else
 #define HAVE_ALIGNED_ACCESS_REQUIRED 1
 #endif
+
+#define HAVE_EXPLICIT_BZERO 1