switch from graph_objects to px

- Update Python 3.6 → 3.11 in runtime.txt
- Create pyproject.toml with modern project metadata
- Update all plotting libraries to latest versions:
  - plotly 4.10 → 5.24+ (no authentication needed)
  - plotnine → 0.13+ (better ggplot2 compatibility)
  - altair → 5.0+ (major version bump)
  - statsmodels 0.11 → 0.14+ (security fixes)
  - seaborn, pandas, matplotlib to latest
- Remove deprecated packages:
  - altair-saver (use native rendering)
  - chart-studio (not needed with plotly v5)
  - duplicate selenium entry
- Update rpy2 to 3.5+ for Python 3.11 compatibility

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Create setup_r.sh to install required R packages (ggplot2, mgcv)
without requiring conda. Works with system R installations via
Homebrew, apt, or CRAN.

This allows using uv for Python dependencies while maintaining
R integration through rpy2 with a separate R installation.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

- Add 'setup' target to install dependencies with uv
- Update 'dev_environment' to use new setup process
- Remove 'plotly_auth' target (no longer needed with plotly v5)
- Update .PHONY declarations
- Add informative messages about conda → uv migration

Core build targets (render, qrender, test) remain unchanged.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Add .github/workflows/deploy.yml with:
- R 4.3 setup via r-lib/actions
- Python 3.11 setup
- uv for dependency installation
- Firefox/geckodriver for selenium
- xvfb for headless rendering
- pytest test execution
- Netlify deployment:
  - Production: master branch
  - Preview: other branches

Required GitHub secrets:
- NETLIFY_AUTH_TOKEN
- NETLIFY_SITE_ID

This modernizes CI from deprecated Travis CI to GitHub Actions
with improved caching and faster builds.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Mark .travis/authenticate_plotly.py as deprecated with explanation.

With plotly v5+, credentials are not required for local rendering.
Plotly authentication is only needed for uploading to Chart Studio.
This project renders plots locally to PNG using Kaleido, so no
authentication is necessary.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

README.md:
- Add modern development setup instructions with uv
- Document system R installation requirements
- Add build commands and CI/CD information
- Clarify plotly no longer needs authentication

CLAUDE.md:
- Update all dependency versions and requirements
- Document new uv-based setup process
- Add comprehensive CI/CD section
- Update technical constraints and configuration files
- Add GitHub Actions workflow documentation

Both files now reflect Python 3.11+, uv package manager,
and GitHub Actions deployment to Netlify.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

MIGRATION_NOTES.md:
- Comprehensive overview of all changes made
- Detailed dependency update table
- Next steps and setup instructions
- Potential issues and rollback plan
- Success criteria checklist

TESTING_CHECKLIST.md:
- Environment setup verification steps
- Build testing procedures
- Known migration issues to watch for
- CI/CD testing guide
- Performance check recommendations

These documents provide guidance for the first build with
modernized dependencies and help identify compatibility issues.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Change default 'all' target from 'render s3_upload' to just 'render'
since deployment is now handled by Netlify via GitHub Actions.

- 'make' or 'make all' now just builds the site
- 'make deploy' added for manual S3 upload if needed
- S3 upload is legacy, kept for backward compatibility

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Use Make's dependency system to automatically install dependencies
before building. This follows standard Make patterns:

- Create .setup_done sentinel file after successful setup
- Make render and run_nb depend on .setup_done
- Running 'make' will now automatically run setup if needed
- Setup only runs once (or after 'make clean')
- Add .setup_done to .gitignore

Now users can just run 'make' and everything works.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Replace `uv pip install -r requirements.txt` with `uv sync` to use
modern uv workflow with pyproject.toml:

- Use `uv sync` to create venv and install dependencies
- Use `uv run` prefix for all Python/Jupyter commands
- Remove build-system from pyproject.toml (not a package)
- Fix deprecation warning (tool.uv.dev-dependencies → dependency-groups)
- Add .venv/ and uv.lock to .gitignore

Benefits:
- Avoids "externally managed" Python installation errors
- Uses proper virtual environment isolation
- Follows modern uv best practices
- Automatic venv creation and management

Tested: `make` now successfully creates venv and installs all Python
dependencies. Only requires system R to be installed separately.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Replace deprecated 'size' parameter with 'height' in seaborn calls:
- sns.lmplot(..., size=12) → height=12
- FacetGrid(..., size=10) → height=10

The 'size' parameter was renamed to 'height' in seaborn 0.11+
and removed entirely in 0.13+.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Seaborn fixes (0.13+ compatibility):
- Fix remaining FacetGrid size -> height parameters (3 instances)
- Update all .pipe(sns.FacetGrid, size=X) calls

Altair fixes (5.0+ compatibility):
- Add vl-convert-python dependency for PNG export
- Update Altair renderer configuration to use 'png' mode with vl-convert
- Remove deprecated selenium-based PNG rendering

These changes ensure all plotting libraries work with their 2025 versions.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Kaleido v1+ requires Chrome for image rendering. Add automatic
Chrome installation to setup process:

- Update Makefile to run kaleido_get_chrome after uv sync
- Update GitHub Actions workflow to install Chrome
- Update pyproject.toml to note Kaleido v1+ requirement
- Update requirements.txt

This ensures Plotly static image export works with Kaleido v1.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>