We now have our own shared memory implementation.

Allows queueing of multiple messages on the TCP socket, which considerably improves throughput.

Sends a list of all queued events in the reply instead of sending them one-by-one. This way, we avoid some communication overhead.

We currently need an additional request to prepare zero-copy shared memory messages. For small messages, it might be faster to avoid this extra round-trip and send the message directly over TCP. This commit implements support for this. Messages smaller than a threshold (currently set to 4096 bytes) are sent via TCP, while larger messages still use shared memory.

This step also enables future optimizations such as queueing output messages in order to improve the throughput.

This commit enables users to pass environment variable as composition
of env variable such as:

```yaml
  - id: yolov5
    operator:
      outputs:
        - bbox
        - ready
      inputs:
        image: oasis_agent/image
        check: oasis_agent/yolov5_check
      python: ../../operators/yolov5_op.py
    env:
      YOLOV5_PATH: $HOME/Documents/dependencies/yolov5
      YOLOV5_WEIGHT_PATH: $DORA_DEP_HOME/dependencies/yolov5/yolov5n.pt
```

Note that this is only for the env field. Expanding env variable for
source path might conflict with using url as source path.

Treat them like normal input IDs with a slash in their name in most cases.

We cannot wait until all operators of the runtime node are finished because operators might be subscribed to other operators.

To allow for individual testing and benchmarking. This will also enable reuse of the library in other projects.

The goal is to avoid the latency of the TCP socket.

Throw an error if a daemon cannot reach the coordinator anymore.

This is only an interim solution. In the future, we want to make the daemon more robust and ideally even allow restarts of the coordinator.

The shared memory region size might be larger because of padding.

Makes the `dora list` command work.