improve pytest of rdt-1b

Add main into rdt-1b

add small cloud fix for rdt 1b

Small rdt-1b main fix

Small improvement on rdt 1b

Small fixes to dora-rdt-1b main

Add piper example

Add environment variable for configuration vision and language parameter

add python feature flag to dora-rerun

Fix play inference

fixing replay issue

make data dir dependant on the date

Make `dora-message` a dependency of `dora-core`, instead of the other way around. This way, we can continue to freely bump the version of `dora-core` with the other workspace crates, without introducing errors such as #708.

In order to get the best performance from Qwenvl2, we need to temporary
add this piece of code that is going to unqueue qwenvl queue.

This is directly linked to: https://github.com/dora-rs/dora/pull/652
which is that we currently having a queue issue within nodes for
`queue_size: 1` where the queue is actually pending older data and not
the newest one.

You will find in this pull request a quick implementation of an openAI
server makes it possible to make request from an openai endpoint client
and receive it within the dataflow.

Note that I have only implemented chat completion and this is still
experimental.

Get started with:

```bash
cd examples/openai-server
dora build dataflow.yml
dora start dataflow.yml

# In a separate terminal
python openai_api_client.py
```

Hi! I took a look at our `examples` directory, and while reading through
some dataflows, I noticed that there are a lot of outdated node
syntaxes:

```YML
nodes:
  - id: rust-node
    custom:
      build: cargo build -p benchmark-example-node --release
      source: ../../target/release/benchmark-example-node
```

I think it would be better if we avoided using deprecated syntax in our
example folder, as it will be one of the first things people look at.