Ploomber

- `DAGSpec` warns if parameter declared in env but unused
- Implements `{SQLDump, NotebookRunner, PythonCallable}.load()`
- `File.client` downloads during task execution instead of render
- Adds `ploomber.OnlineModel`, which provides a simpler API than `OnlineDAG` for models that implement a `.predict()` method
- Adds function to find package name if using standard layout

- Changes `extract_product` default in spec API to False
- Tasks get a default name equal to the filename without extension (e.g., plot.py -> plot)
- `File` saves metadata in a `.{filename}.metadata` file instead of `{filename}.source`
- Adds `ploomber examples` command
- Adds Deployment guide to documentation
- `EnvDict` loads `env.yaml` and uses values as defaults when passing a custom dict
- Simpler repr for SQL products
- Improved Spec API docs
- Adds `ploomber.tasks.TaskGroup.from_params` to create multiple tasks at once

- Changes a lot of error messages for clarity
- Clearer `__repr__` for `Placeholder`, `File`, and `MetaProduct`
- Default placeholders can be used in `pipeline.yaml` without defining `env.yaml`
- Better formatting for displaying DAG build and render errors
- Spec API initializes task spec as `SQLDump` if product has suffix `.csv` or `.parquet`
- Coloring CLI error traceback
- Spec API skips `SourceLoader` if passing an absolute path
- `DAG.clients` validates keys (using `DAGClients`)
- `params` available as hook argument
- Rewritten Spec API documentation

- Better display of errors when building or rendering a DAG (layout and colors)
- `File` implements the `os.PathLike` interface (this works now: `pandas.read_parquet(File('file.parquet'))`)
- Several error messages refactored for clarity
- Adds `DAGSpec.find()` to automatically find `pipeline.yaml`

- Adds `OnlineDAG` to convert `DAG` objects for in-memory inference
- Spec API (`pipeline.yaml`) supports DAG-level and Task-level `serializer` and `serializer`
- CLI looks for `src/{pkg}/pipeline.yaml` if `pipeline.yaml` doesn't exist
- Adds `{{cwd}}` placeholder for `env.yaml` that expands to current working directory

- Support for Python 3.9
- `SQLAlchemyClient` now accepts an argument to pass custom parameters to `sqlalchemy.create_engine`
- Temporarily pins papermill version due to an incompatibility with jupytext and nbformat (jupytext does not support cell ids yet)
- Adds `--on-finish/-of` to `ploomber task` to execute the `on_finish` hook
- DAGs with R notebooks can render even if the ir kernel is not installed