Support serde flatten annotation for schema derivation and de/serialization

[Sufflope]

Fixes #247

[Kriskras99]

I think this is a good option for implementing support for #[serde(flatten)].
The one thing I don't like, is that we store the the flatten boolean in the custom attributes of RecordSchema. However, the only alternative I see, is adding another field to that struct which would be a breaking change.

@martin-g what do you think?

[martin-g]

@martin-g what do you think?

Honestly, I don't want to maintain anything Serde related :-|
Every time I tried to dive deep into Serde related code I see hacks and functionalities contradicting with each other.
I am not smart enough to understand why the serde crate is so popular...

[Kriskras99]

Serde is popular because it's really easy to use as a user. The complicated parts are hidden in serde itself and any crate implementing formats.

@Sufflope in the async work I'm also trying to make the serde part easier to maintain, I'll get back to this PR after the async work.

[Kriskras99]

I think this is the wrong approach. I think the flattened fields should be included in the derived schema directly, so that the writer doesn't need to add special support for this.

Also, correct me if I'm wrong, but this PR only works for people using Writer::append_ser not those using to_value?

[Kriskras99]

@Sufflope thank you for your PR (and your temporary serdavro crate)!. We have a more maintainable and complete implementation in #359, so I'll close this PR in favour of that one.
Other improvements to the Serde/Avro integration are tracked in #361, if you run into more problems with Serde/Avro integration please let us know!