Fix Semantic Layer for Fusion metric discovery

[C00ldudeNoonan]

Summary

• migrate legacy top-level Semantic Layer YAML into Fusion-native model-owned semantic metadata
• move simple metrics inline with their source models and keep metric names discoverable via dbt Fusion
• remove legacy semantic model, metric, and saved query files that block Dagster/dbt Core CI parsing

Fixes #101

Validation

• dbt parse
• dbt sl list metrics --project-dir /Users/alexnoonan/code/econ-data-platform/economic-data-project-full-wt-issue-101/dbt_project
• cd macro_agents && uv sync --extra dev && uv run ruff check . --exclude .claude && uv run ruff format --check . --exclude .claude && uv run ty check --extra-search-path src src/macro_agents && BIGQUERY_PROJECT=econ-data-project-478800 DBT_ORCHESTRATION_MODE=local uv run dg check defs && uv run pytest tests/ -v -m 'not skip_ci'

Notes

• dbt Core in Dagster CI does not register Fusion-inline metrics for legacy saved query dependencies, so saved queries are removed in this PR. Metrics remain discoverable through Fusion.