Add subdirectory support for hf:// URLs and gs:// scheme for GCS

[baogorek]

Summary

• Add subdirectory support for hf:// URLs (e.g., hf://owner/repo/path/to/file.h5)
• Add gs:// URL scheme for Google Cloud Storage (e.g., gs://bucket/path/to/file.h5)

Changes

hf:// improvements

• Add parse_hf_url() helper function to centralize URL parsing
• Support subdirectory paths: hf://owner/repo/path/to/file.h5[@version]
• Fix inconsistent parsing between dataset.py and simulation.py

gs:// new functionality

• Add google_cloud.py module with parse_gs_url(), download_gcs_file(), upload_gcs_file()
• Support URLs: gs://bucket/path/to/file.h5[@version]
• google-cloud-storage is optional - raises helpful ImportError if not installed

Test plan

• 6 new tests for parse_hf_url()
• 10 new tests for parse_gs_url()
• All 449 existing tests pass

Closes #405

🤖 Generated with Claude Code

[baogorek]

I just tested this manually as well:

In [1]: from policyengine_us import Microsimulation

In [2]: sim = Microsimulation(dataset="hf://policyengine/test/states/RI.h5")

In [3]: sim.calculate('household_id').weights.sum()
Out[3]: np.float64(393240.457935619)

In [4]: sim = Microsimulation(dataset="gs://policyengine-us-data/states/RI.h5")

In [5]: sim.calculate('household_id').weights.sum()
Out[5]: np.float64(393240.457935619)

[anth-volk]

Thanks for this @baogorek! Tested locally and found no issues.