[codex] Add experimental MCP rollout and demo notebook

[cschrupp]

Summary

This PR adds the experimental MCP rollout for wellplot and the accompanying developer notebook demo.

It includes:

• the stdio-first wellplot-mcp server with tool, resource, and prompt registration
• service-layer support for scoped previews, example export, logfile text validation/format/save, and explicit file rendering
• packaged production example assets for MCP resources
• MCP tests, installed-wheel smoke coverage, and release-workflow verification for the optional MCP extra
• version bump and changelog/release-process updates for the 0.3.0 rollout
• public docs for MCP install, workflow, and API reference
• a generated developer notebook and helper script demonstrating the current MCP capabilities

Why

The project already had YAML and Python API workflows, but it did not yet expose a supported MCP surface for agent-driven logfile review and authoring. This PR ships that surface as an experimental feature and updates the release/docs path so it can be published coherently.

The notebook/demo work also makes the current MCP behavior easier to inspect before release.

User and developer impact

Users can now:

• install wellplot[mcp]
• launch wellplot-mcp
• validate, inspect, preview, and render logfiles through MCP
• export packaged example bundles
• validate, normalize, and save full logfile YAML text safely under the server root

Maintainers also get:

• release verification for the MCP extra in a clean environment
• MCP documentation wired into the published docs site
• a developer notebook that exercises the current MCP surface end to end

Validation

Local checks run for this branch included:

• uv run ruff check examples/mcp_workflow_demo.py scripts/generate_example_notebooks.py src/wellplot/mcp tests/test_mcp_service.py tests/test_mcp_server.py
• uv run pytest tests/test_mcp_service.py tests/test_mcp_server.py -q
• uv run python scripts/generate_example_notebooks.py --check
• uv run --group docs mkdocs build --strict

Notes

• MCP remains experimental in this PR.
• PyPI publication is still manual through the existing Release workflow.
• The generated user notebooks changed only because their embedded version output now reflects 0.3.0.