Dependency Pruning, Version Pin Review, and Conda Removal

[i-am-sijia]

Following the implementation of dependency locking for ActivitySim using UV in Phase 10 (PR #956), several follow-up items were identified that were out of scope for that phase but are important for future maintenance and simplification of the environment. @josiekre

Tasks to Consider

• [ ] Remove unused dependencies.
  • The initial implementation (PR #956) retained all dependencies from the legacy Conda environment YAML files.
  • Some unused packages (e.g., orca), were removed, but a complete cleanup requires careful auditing and testing, which was beyond the Phase 10 scope.
• [ ] Remove version pins in pyproject.toml, e.g., [1][2].
  • With UV, we should avoid manually pinning dependency versions with upper bounds (lower bounds are acceptable).
  • If a pin is absolutely necessary, it should be tracked via a GitHub issue and regularly re-evaluated.
  • We should also aim to include only direct dependencies in pyproject.toml.
• [ ] Fix Sharrow test failures on Mac and Linux
  • Some Sharrow tests were failing on Mac and Linux [1], although all of them pass on Windows.
• [ ] Update benchmarking functions
  • These functions require code changes to remove Conda dependencies [1], or we should consider deprecating them.
• [ ] Remove conda references from all Jupyer notebooks in the repository.
  • E.g., [1][2] Most references are just instructional comments, but they should be cleaned up for consistency.