LibCrypto: Use explicit_bzero instead of memset to zero 'secure data'

PVS-Studio flagged this, as memset can be optimized away by the compiler
in some cases. We obviously don't want that to ever happen so make sure
to always use `explicit_bzero(..)` which can't be optimized away.

Userland/Libraries/LibCrypto/Hash/MD5.cpp

@@ -6,6 +6,7 @@
 
 #include <AK/Types.h>
 #include <LibCrypto/Hash/MD5.h>
+#include <string.h>
 
 static constexpr u32 F(u32 x, u32 y, u32 z) { return (x & y) | ((~x) & z); };
 static constexpr u32 G(u32 x, u32 y, u32 z) { return (x & z) | ((~z) & y); };
@@ -199,7 +200,7 @@ void MD5::transform(const u8* block)
     m_C += c;
     m_D += d;
 
-    __builtin_memset(x, 0, sizeof(x));
+    explicit_bzero(x, sizeof(x));
 }
 
 }

Userland/Libraries/LibCrypto/Hash/SHA1.cpp

@@ -7,6 +7,7 @@
 #include <AK/Endian.h>
 #include <AK/Types.h>
 #include <LibCrypto/Hash/SHA1.h>
+#include <string.h>
 
 namespace Crypto {
 namespace Hash {
@@ -63,7 +64,7 @@ inline void SHA1::transform(const u8* data)
     c = 0;
     d = 0;
     e = 0;
-    __builtin_memset(blocks, 0, 16 * sizeof(u32));
+    explicit_bzero(blocks, 16 * sizeof(u32));
 }
 
 void SHA1::update(const u8* message, size_t length)