[http server] Batch requests (#292)

* WIP

* Implement draft of batch requests

* fmt

* cleanup

* Explain why we don't use an untagged enum

* Avoid allocating a Vec for single requets

* Add comment

* Add a benchmark for batch requests

* Add more tests, noting where we diverge from the spec
Fix empty batch case, i.e. `[]`

* Obey the fmt

* Update benches/bench.rs

Co-authored-by: Andrew Plaza <[email protected]>

* Update http-server/src/server.rs

Co-authored-by: Andrew Plaza <[email protected]>

* Add link to issue

* Explain why we're closing the receiving end of the channel.

* Limit logging of requests and response to 1kb
Add more comments
Factor out batch response collection

* Wrap comment

* tweak log line

* Benchmark batch request over different batch sizes

* fmt

Co-authored-by: Andrew Plaza <[email protected]>