Feat/visualization module

[birsenyrtds]

Summary

This PR adds a modular visualization stack under src/visualization/:

• flexvis.py: Strategy Registry + Policy architecture for heatmaps and embeddings (UMAP/PCA), including normalization, ordering/clustering, colormap scaling, and multiple plotting backends (matplotlib/plotly/plotly.express).
• plot.py: A Dash-based demo UI that renders a heatmap and a UMAP plot with interactive controls.

Also adds requirements.txt with the required dependencies: dash, plotly, umap-learn, scikit-learn, matplotlib.

What’s included?

• CombinedPolicy to run heatmap + UMAP in a single call.
• Pluggable backends:
  • Heatmap: matplotlib or plotly.
  • UMAP scatter: plotly.express, plotly, or matplotlib.
• Parametric UMAP (via policy kwargs): n_neighbors, min_dist, metric, random_state, etc.
• Optional pre-UMAP PCA (exposed via a Dash slider) to denoise and improve cluster separation.
• Extensible Strategy Registry: add custom normalizers/clusterers/cmap scales/reducers/plotters by simple registration.

Why?

• Reduces duplicated code by centralizing visualization logic.
• Provides a clean, extensible architecture for adding new methods.
• The Dash UI enables quick iteration and fast visual feedback for parameter tuning.

How to run

pip install -r requirements.txt
python -m src.visualization.plot
# If the browser doesn't open automatically:
# http://127.0.0.1:8050