Establish Arch data foundation

[MaxGhenis]

Summary

• Rename the project surface from microplex-sources to Cosilico Arch (cosilico-arch, arch CLI, arch Python namespace).
• Add Arch source/fact/normalization/target-input modules with lineage-preserving source facts.
• Reframe calibration data as Arch target inputs while leaving source selection, reconciliation, aging, and active target sets to Microplex Targets.
• Update SNAP and ACA ETLs to build target inputs through source facts and representation-only normalization helpers.
• Remove legacy target variable compatibility helpers and stale wrapper modules.
• Address review findings by removing the local microplex package collision, moving the old pipeline entry point under arch.pipelines, making Supabase schema boundaries explicit, and adding a wheel-install smoke test to CI.

Validation

• git diff --check
• uv run ruff check arch db/supabase_client.py db/etl_snap.py db/etl_aca_enrollment.py db/schema.py calibration/constraints.py calibration/targets.py micro/us/district.py tests/test_arch_facts.py tests/test_arch_namespace.py tests/test_arch_normalization.py tests/test_etl_snap.py tests/test_etl_aca_enrollment.py
• uv run pytest tests/test_arch_facts.py tests/test_arch_namespace.py tests/test_arch_normalization.py tests/test_etl_snap.py tests/test_etl_aca_enrollment.py tests/test_calibration.py -q
• uv build
• Clean wheel install smoke test: import arch, import arch.targets, from micro.us.district import DistrictMicroplex, from microplex import ConditionalMAF, and arch --help

Notes

The repository is temporarily named CosilicoAI/arch-data because GitHub has retired the CosilicoAI/arch namespace. A support request has been sent asking GitHub to release that slug.