Generate schema v2 catalog fields from ddp sync catalog

Parent epic: #15

Depends on: #16, #21

Why

The catalog must remain generated, not hand-edited. Today ddp sync catalog renders schema v1 from package metadata, installed data_designer.plugins entry points, and direct data-designer dependency constraints. It needs to render the concrete schema v2 contract from #16 using the repo-level tap metadata from #21.

Implementation

• Set CATALOG_SCHEMA_VERSION = 2.
• Extend CatalogEntry with:
  • source: dict[str, object]
  • docs_url: str
• Generate default NVIDIA entries with this source object because [tool.ddp.tap].default-source = "pypi":

{"type": "pypi", "package": "<project.name>"}

• Generate docs.url as:

f"{docs_base_url.rstrip('/')}/plugins/{normalize_docs_slug(project_name)}/"

where normalize_docs_slug should match the existing plugin docs slug logic.

• Preserve current v1 fields exactly unless #16 says otherwise: name, plugin_type, description, package, entry_point, and compatibility.
• Keep compatibility.data_designer.requirement, specifier, and marker; the PDF example omits requirement, but the current field is useful for explanations and should remain.
• Add duplicate runtime plugin-name detection before rendering JSON.
• Add source-object validation for the default pypi source.
• Regenerate catalog/plugins.json.
• Update devtools/ddp/tests/test_catalog.py expected output to schema v2.
• Add a multi-entry fixture test where two entry points in one package share identical package and source metadata.

Expected generated entry

For the template plugin, schema v2 output should be structurally equivalent to:

{
  "name": "text-transform",
  "plugin_type": "column-generator",
  "description": "Template Data Designer plugin — text transform column generator",
  "package": {
    "name": "data-designer-template",
    "version": "0.1.0",
    "path": "plugins/data-designer-template"
  },
  "entry_point": {
    "group": "data_designer.plugins",
    "name": "text-transform",
    "value": "data_designer_template.plugin:plugin"
  },
  "compatibility": {
    "python": {"specifier": ">=3.10"},
    "data_designer": {
      "requirement": "data-designer>=0.5.7",
      "specifier": ">=0.5.7",
      "marker": null
    }
  },
  "source": {
    "type": "pypi",
    "package": "data-designer-template"
  },
  "docs": {
    "url": "https://nvidia-nemo.github.io/DataDesignerPlugins/plugins/data-designer-template/"
  }
}

Acceptance criteria

• make catalog emits schema_version: 2.
• Every generated plugin entry has source.type = "pypi", source.package = package.name, and a docs.url derived from [tool.ddp.tap].docs-base-url.
• Existing compatibility fields remain generated from direct versioned data-designer dependencies.
• Duplicate runtime plugin names fail catalog generation.
• Multi-entry packages render as multiple entries sharing one package/source identity.
• uv run pytest devtools/ddp/tests/test_catalog.py -q passes.
• make check-catalog fails when checked-in catalog output is stale.

Dependencies

• Depends on: #16, #21.
• Blocks: #18, #19, #20, #23, and DataDesigner CLI read-only catalog work.

Start by reading the existing catalog generator behind ddp sync catalog and devtools/ddp/tests/test_catalog.py, including the current plugin documentation slug logic. Run uv run pytest devtools/ddp/tests/test_catalog.py -q and inspect catalog/plugins.json; done means schema version 2 output, validated source and docs fields, duplicate-name failure, multi-entry coverage, and passing catalog checks.