Motivation:

There were a couple of easy performance wins in NIOAsyncChannel which cut the allocations in half again. Those performance wins are mostly around the underlying NIOAsyncWriter and NIOAsyncSequenceProducer.

Modifications:

This PR contains a few changes which split up into separate commits to make reviewing easier:

1. I added a fast path to the NIOAsyncChannelInboundStreamChannelHandlerProducerDelegate which allows it to skip the thread hop when we are already on the correct EventLoop.
2. I copied over the _TinyArray implementation from swift-certificates to optimise the fast path of having just a single producer suspended in the NIOAsyncWriter.
3. I implemented the NIOAsyncWriterSinkDelegate customization point for single element writes in the NIOAsyncChannelOutboundWriterHandler. This avoids allocating a Deque for single element writes.
4. I made sure that we call the single element customization point more often in NIOAsyncWriter and I moved the storage of the suspendedYields to the _TinyArray which optimises for the single producer case.

Result:

Halved the allocations for the NIOAsyncChannel benchmark again.