Fix Codec T64 on s390x

src/Compression/CompressionCodecT64.cpp

@@ -307,7 +307,19 @@ void reverseTransposeBytes(const UInt64 * matrix, UInt32 col, T & value)
 template <typename T>
 void load(const char * src, T * buf, UInt32 tail = 64)
 {
-    memcpy(buf, src, tail * sizeof(T));
+    if constexpr (std::endian::native == std::endian::little)
+    {
+        memcpy(buf, src, tail * sizeof(T));
+    }
+    else
+    {
+        /// Since the algorithm uses little-endian integers, data is loaded
+        /// as little-endian types on big-endian machine(s390x, etc.)
+        for (UInt32 i = 0; i < tail; i++)
+        {
+            buf[i] = unalignedLoadLE<T>(src + i * sizeof(T));
+        }
+    }
 }
 
 template <typename T>