docs: architecture/sequence/module diagrams, real CLI screenshots, badges, CI

[codeapplied]

Visual pass across the repo, per Joy's request to add visuals wherever a reviewer would actually look.

• README.md: badges (tests, Python version, license), a Mermaid architecture/data-flow diagram, and an 'In action' section with 3 real CLI-output SVG images (run --apply, health, reconcile) — genuinely captured Rich console output, not mockups.
• docs/ARCHITECTURE.md: the ASCII data-flow diagram replaced with a fuller Mermaid version, a new Mermaid sequence diagram of one daily run, and a new Mermaid module-relationship diagram.
• .github/workflows/tests.yml: new — runs the test suite on every push/PR to main. Needed for a legitimate 'tests passing' badge; previously only the scheduled pipeline-runner workflow existed, nothing actually ran CI.
• scripts/generate_readme_assets.py: committed (not thrown away) so the CLI-output SVGs can be regenerated after any future CLI change, rather than going stale.
• GitHub Topics: added (python, web-scraping, crm-integration, automation, sqlalchemy, typer, microsoft-graph-api, pipedrive) for discoverability — done directly via the API, not part of this PR's diff.

Verified: all 55 tests still pass; the new tests.yml and existing daily-fetch.yml both validated as syntactically correct YAML; the asset-generation script actually run end-to-end (confirmed real 5/1/4 fetch/new/filtered counts and a genuine DB-vs-Excel discrepancy in the reconcile screenshot, not staged data).

One thing I could not verify myself: Mermaid diagram rendering. I applied defensive quoting throughout (any label with parens, arrows, or special characters) since a few of GitHub's Mermaid parsing edge cases aren't something I can render-test from here — please check the 4 diagrams actually render correctly once this PR is viewable on GitHub before merging.