Fix splice optimization

[rchernobelskiy]

The intended behavior from the comment is not taking place:

// Unwrap the src and dst from *Conn to *net.TCPConn so Go
// 1.11's splice optimization kicks in.
src = UnderlyingConn(src)
dst = UnderlyingConn(dst)

The UnderlyingConn func doesn't return a *net.TCPConn so I think the intention here was to call tcpConn instead.