Performance of stdin -> stdout Baseline Degrades the more Worker Threads are Added

I found this when looking into #7935 with @jakelandis and it looks like we have some more room for improvement here.
Currently for a config like

```
input { stdin{}} output {stdout {codec => "dots"}}'
```

you can clearly observe a decrease in throughput the more workers you add. Eventually, you simply run into the `if` part of the synchronous queue read here, which is very slow:

```rb
def read_next
read_size = @queue.queue.drainTo(@originals, @size)
if read_size < @size
(@size - read_size).times do |_|
e = @queue.poll(@wait)
return if e.nil?
@originals.add(e)
end
end
end
```

Just a suggestion:
We could probably way improve this behavior if we enqueued as full as possible full batches on the writer side (instead of individual elements) and simply pulled those off the Queue one by one.
This means that the `@wait` would move to the writer side and we'd only have a `take()` on the reader side that returns a batch of events (could be an array).