Merge pull request #8749 from ikopylov/function-bitCount-Int16-Int8-perf

bitCount performance: special handling for Int16 and Int8

dbms/src/Functions/bitCount.cpp

@@ -20,6 +20,10 @@ struct BitCountImpl
             return __builtin_popcountll(a);
         if constexpr (std::is_same_v<A, UInt32> || std::is_same_v<A, Int32> || std::is_unsigned_v<A>)
             return __builtin_popcount(a);
+        if constexpr (std::is_same_v<A, Int16>)
+            return __builtin_popcount(static_cast<UInt16>(a));
+        if constexpr (std::is_same_v<A, Int8>)
+            return __builtin_popcount(static_cast<UInt8>(a));
         else
             return __builtin_popcountll(ext::bit_cast<unsigned long long>(a));
     }