Stream-infer

What's Changed

- Fix DynamicApp parameters are not accurate
- `args` -> `kwargs`
- add `args` tuple
- Fix some typing and hints

What's Changed

- fix(package): move the project source code from `src`
- feat(Dynamic): add dynamic import `process` and set `recording_path`
- chore: update examples & README.md

What't Changed

- **💥Broken** refactor(typo): specification of variable names adds typing
- **💥Broken** refactor(Dispatcher): streamline the Dispatcher implementation, remove collect_results and move to DevelopDispatcher
- feat: added `DynamicApp` to support dynamically running inference tasks through configuration
- typo(Streamlit): rename the streamlit app file to streamlit_.py instead of server.py
- chore: update examples

What's Changed

- chore(Dispatcher): Set the default `max_size` of `Dispatcher` to 30
- feat(Player): Added `tqdm` for limited length video progress printing
- feat(Timer): The timeout warning is added to the inference Timer in real-time inference
- feat(Dispatcher): Run the Inference algo `frame_step` set to 0

What's Changed

- feat(Dispatcher): using the factory mode simplifies instantiation of the Dispatcher
- feat(Recorder): supports offline inference to record video files to disk according to different producers
- package: separate version

- Fix(Player): number of frames under real-time inference does not correspond to time
- Feat(Server): support for loading live streams and producer selection