feat: refine template manifesto with use case tags

[DaveHanns]

Summary

Refines the useCases tags across Actor templates so the Console wizard's "use case + language" filter produces a meaningful split, polishes a handful of template descriptions, and adds in-place documentation for what each use case represents.

Changes

useCases tag refinements

• python/ts/js-empty: [WEB_SCRAPING] → [STARTER] (empties are topic-agnostic, not scrapers).
• python/ts/js-standby (×3): added INTEGRATION to [STARTER] (Standby actors serve HTTP, callable by outside systems).
• python-mcp-proxy, python-mcp-empty, ts-mcp-proxy, ts-mcp-empty (×4): added INTEGRATION to [AI] (callable by AI clients via the MCP protocol).
• ts-playwright-test-runner: [WEB_SCRAPING] → [AUTOMATION] (it's a test runner — automation, not scraping).
• python-playwright / python-selenium: added AUTOMATION to [WEB_SCRAPING] (pure browser tooling without a scraping framework — repurposable for automation).
• cli-start: filled in the previously missing useCases with [STARTER, INTEGRATION] (the Dockerfile sets up ubi to install third-party binaries — Actorization is integration).

Description polish

• Rewritten descriptions for 9 templates (python-crawlee-beautifulsoup, python-empty, python-standby, python-crawlee-parsel, python-crawlee-playwright, python-crawlee-playwright-camoufox, python-smolagents, js-langchain, cli-start) — clearer, more direct phrasing, and consistent grammar (e.g. "Start from a given start URLs" → "Start from a URL").

Documentation

• Added _useCaseDescriptions field at the manifest top-level — a self-contained reference for what each tag represents. Canonical enum + JSDoc lives in apify-core (TEMPLATE_USE_CASES in src/packages/consts/src/templates.ts).

Tagging principles applied

The tagging principle is: tag for what the template IS (its primary framing/purpose), not what it COULD become. Concretely:

• Empty scaffolds carry only STARTER (no inherent topic).
• STARTER is for entry-level / first-Actor scaffolds. Skeletons explicitly aimed at experienced devs don't qualify.
• Pure browser tooling (raw Playwright / Selenium, no Crawlee wrapping) can earn AUTOMATION in addition to WEB_SCRAPING — the underlying tool is genuinely repurposable.
• Crawlee variants stay WEB_SCRAPING only — Crawlee's crawler-specific abstractions (request queue, *Crawler classes) orient the project toward scraping, even when paired with Playwright/Puppeteer.
• AI agents stay AI-only, even when they crawl as part of their workflow. Calling Apify Actors as tools or fetching web data is a means; the AI framework demo is the product.
• INTEGRATION is reserved for templates that meaningfully bridge Apify with external systems: Standby HTTP servers, MCP servers callable by AI clients, or third-party-tool wrappers (cli-start). Not "any Actor that has an API surface" — that would be everything.

Coverage after changes

Use case → templates

Use case	Count
WEB_SCRAPING	21
STARTER	17
AI	13
INTEGRATION	8
AUTOMATION	4

Use case × language

Use case	JavaScript	Python	TypeScript	Total
WEB_SCRAPING	6	9	6	21
STARTER	5	7	5	17
AI	2	7	4	13
INTEGRATION	2	3	3	8
AUTOMATION	1	2	1	4

Open for refinement

This pass focused on obvious mistags and missing tags. Further iteration welcome — especially around:

• JS / TS AUTOMATION = 1 each — js-cypress and ts-playwright-test-runner respectively. Both are test runners. No general-purpose JS/TS browser-automation template exists; an equivalent of python-playwright/python-selenium for each would fill the gap.
• JS INTEGRATION = 2 — js-standby + cli-start. No JS MCP template exists; adding one would mirror the Python/TS coverage.
• Individual template judgement calls — tagging is partly subjective. Happy to iterate on any specific template.

No changes to template content or behavior — only metadata in manifest.json.