Help improve experience with untagged schemas

[epage]

I expect a lot of users are like me and just assume their code is fast enough. It wasn't until I had to do more explicit benchmarks that I found out how slow untagged is. The hope is to make this more discoverable for others along with an easier / less intimidating way of improving performacne via serde-untagged. See also serde-rs/serde#2101

Similarly, errors can be bad and expecting wasn't documented until #153. This also adds cross-linking to make this attribute more discoverable.