Clean up Causal Analysis of Success in Modern Society Tutorial #472

[bkk8403]

Related to #472

Summary

This PR brings the Causal Analysis of Success tutorial into compliance with
the tutorials_checklist.md and X_in_60_mins.format_rules.md standards.

Notebooks were renamed from underscore to dot convention so that
causal_success_API becomes causal_success.API and causal_success_example
becomes causal_success.example, matching the naming pattern used across all
other tutorials. All internal cross-references were updated accordingly.

The Dockerfile was rewritten against the python:3.12-slim template from
class_project/project_template, replacing the previous ubuntu:24.04 base.
Dependencies in requirements.txt are now pinned and grouped by section.

Heavy logic that was previously defined inline in the notebooks, including
DML fitting, Causal Forest estimation, policy comparison, and a policy
runner helper, has been moved into causal_success_utils.py. The notebooks
now only orchestrate and display results, as required by the format rules.

All markdown across the project, covering the README, the how-to guide, and
every notebook cell, was rewritten as flowing prose. Typos identified across
both notebooks were corrected.

Several missing files were added. An init.py makes the directory a proper
package. A .dockerignore was copied from the project template. A
test/test_docker_all.py was created for Docker-based notebook execution tests.
The conftest.py no longer blanket-disables test collection. Unit tests were
written for the three new utility functions.

The README was trimmed to 34 lines and now includes a Quick Start section and
a changelog. The how-to guide path references were updated to reflect the
actual directory location.

Both notebooks were verified to run end-to-end inside Docker without errors.

Test plan

Build the Docker image with ./docker_build.sh, then launch Jupyter with
./docker_jupyter.sh. Open causal_success.API.ipynb first and run all cells,
which should complete in two to three minutes. Then open
causal_success.example.ipynb and run all cells, which takes five to ten
minutes because of the DML and Causal Forest fitting. Both should produce
output without tracebacks.