Add CI tests

[jqnatividad]

So we can iterate on codebase with confidence that we're not introducing bugs/regressions

[jqnatividad]

@minhajuddin2510 , let's migrate the dp+ testing harness you built to GH Actions, so we can iterate with confidence, and DP+ users can have their own custom test cases on their DP+ forks.

[jqnatividad]

Done

[jqnatividad]

Closing as substantially satisfied across multiple PRs:

• Python unit tests — 192/192 pass on main covering qsv wrappers, the Prefect flow tasks, validation/quarantine, AI suggestions, format conversion + delimiter detection, CLI migration helpers, the dictionary stash + restore, file persistence, metadata stats cleanup, and security. Run inside the dpp-test container (ckan/ckan-dev:2.11 with GDAL + qsv).
• JS unit tests — 12/12 pass via Vitest + jsdom for the scheming-ai-suggestions.js polling state machine + DOM updates (test(js): vitest + jsdom unit tests for scheming-ai-suggestions.js (+ fixes 1 polling bug) #304).
• Integration suite — tests/integration/ runs the full Prefect → Postgres → Solr → Redis stack via scripts/integration-up / docker-compose.integration.yaml (chore(integration): make the integration stack runnable locally end-to-end #300). CI runs it on every PR via .github/workflows/main.yml.
• Roborev + Copilot — every PR in the recent series (feat(metadata): persist dpp_spatial_extent for CSV lat/lon resources #298 → fix(#265): stash Data Dictionary before delete, restore on rollback #307) has been gated on both an LLM-driven code review (roborev) and Copilot's inline review before merge.

Bug-catch evidence the suite is doing real work: PR #303 caught two real production bugs in the qsv describegpt wrapper (wrong flag, wrong envelope shape) via end-to-end with LM Studio; PR #304's JS tests caught a polling bug where the JS read STATUS from the wrong path and would never have terminated; PR #307's tests caught a rollback type-mapping hole that would have silently downgraded numeric columns to text on restore.

If a specific gap remains, please open a focused follow-up. Closing.