R package: use Config/reticulate for Python dependency declaration

Summary

The R package currently manages Python dependencies manually in setup_policyengine() via custom reticulate::virtualenv_install() calls. We should also declare the Python dependency formally using the Config/reticulate field in DESCRIPTION, which is reticulate's standard mechanism.

Current state

• setup_policyengine() manually creates a virtualenv and installs policyengine-taxsim from GitHub
• policyengine_us_version param allows pinning a transitive dep, but this is a custom pattern
• No Config/reticulate field in DESCRIPTION

Proposed change

Add to DESCRIPTION:

Config/reticulate:
  list(
    packages = list(
      list(package = "policyengine-taxsim", pip = TRUE)
    )
  )

Limitations to investigate

• Config/reticulate doesn't natively support git+ URLs — may need to use PyPI once we publish there
• No standard way to pin transitive dependencies (e.g. policyengine-us==1.555.0) via Config/reticulate — the custom policyengine_us_version param may still be needed for that use case
• Need to test interaction with the lazy auto-setup in policyengine_calculate_taxes()

References

• reticulate: Python package installation
• Current implementation: r-package/policyenginetaxsim/R/setup.R

Start with r-package/policyenginetaxsim/R/setup.R and the DESCRIPTION file. Read setup_policyengine() and policyengine_calculate_taxes() to understand the current virtualenv setup and lazy auto-setup, then test how the proposed Config/reticulate declaration interacts with the GitHub dependency and policyengien_us_version. Done means the dependency is declared appropriately and the setup behavior is verified.