Extend ddp new to scaffold all DataDesigner plugin types

[eric-tramel]

Parent epic: #15

Depends on: #21

Why

The runtime plugin system supports column generators, seed readers, and processors. The current scaffold emits only a column-generator package, so DDPlugins is not yet a complete reference tap authoring workflow.

CLI contract

Extend ddp new to:

uv run ddp new my-plugin --type column-generator
uv run ddp new my-plugin --type seed-reader
uv run ddp new my-plugin --type processor

Backwards compatibility:

uv run ddp new my-plugin

must continue to create a column-generator plugin and the help text must say that this is the default.

Generated column-generator template

• Config class subclasses data_designer.config.base.SingleColumnConfig.
• Discriminator field is column_type: Literal["<slug>"] = "<slug>".
• Implementation class subclasses ColumnGeneratorFullColumn[ConfigClass].
• Plugin object uses PluginType.COLUMN_GENERATOR.
• Test imports plugin and calls assert_valid_plugin(plugin).

Generated seed-reader template

• Config class subclasses data_designer.config.seed_source.FileSystemSeedSource.
• Discriminator field is seed_type: Literal["<slug>"] = "<slug>".
• Implementation class subclasses data_designer.engine.resources.seed_reader.FileSystemSeedReader[ConfigClass].
• Generated implementation should be a minimal placeholder that imports cleanly and clearly marks get_dataset_uri / DuckDB connection behavior as TODOs rather than pretending to be production-ready.
• Plugin object uses PluginType.SEED_READER.
• Test imports plugin and calls assert_valid_plugin(plugin).

Generated processor template

• Config class subclasses data_designer.config.base.ProcessorConfig.
• Discriminator field is processor_type: Literal["<slug>"] = "<slug>".
• Implementation class subclasses data_designer.engine.processing.processors.base.Processor[ConfigClass].
• Generated implementation overrides process_after_batch and returns data unchanged with a TODO.
• Plugin object uses PluginType.PROCESSOR.
• Test imports plugin and calls assert_valid_plugin(plugin) if DataDesigner validates processor plugins through that helper; otherwise the test should cover public import and plugin object construction while documenting the DataDesigner validation gap.

Shared generated behavior

• Use tap config from Add repo-level tap metadata config for ddp tooling #21 for package prefix, author, repository URL, docs URL examples, and data-designer dependency.
• Use absolute imports only.
• Use Python 3.10+ annotations.
• Generate README and docs text that says “plugin type” or “entry point,” not “column type,” unless the generated plugin is specifically a column generator.
• Add scaffold golden tests for file tree and important file contents for all three plugin types.
• Add CLI tests for valid types, invalid type errors, and default behavior.

Acceptance criteria

• ddp new my-plugin --type column-generator creates a valid column-generator plugin.
• ddp new my-plugin --type seed-reader creates a valid seed-reader plugin.
• ddp new my-plugin --type processor creates a valid processor plugin or a documented processor-validation-limited scaffold if DataDesigner lacks full validation.
• uv run pytest devtools/ddp/tests/test_cli.py devtools/ddp/tests/test_scaffold.py -q covers all variants.
• Generated docs and plugin cards do not label seed readers or processors as column types.

Dependencies

• Depends on: Add repo-level tap metadata config for ddp tooling #21.
• Related DataDesigner dependency marker: DD-CORE-PROCESSOR-PLUGIN-VALIDATION.
• Blocks: external tap authoring workflow docs in Document the plugin tap contract, workflows, and ecosystem boundaries #25.