docs(skill): add Common pitfalls section (F19+F22+F33+F48+F50+F52+F56)

[DaveHanns]

Summary

Add a Common pitfalls (from real agent runs) section to skills/apify-actor-development/SKILL.md, sitting right before Best practices. Seven concrete teachings distilled from a 14-scenario × 5-stack eval sweep.

Teachings included

Finding	Teaching
F22	Actor version must be MAJOR.MINOR; 1.0.0 fails at platform admission after upload.
F19	Prefer the default dataset for single-shape output; only create named datasets for multi-stream outputs.
F33	Always start from a template; signal for a valid scaffold is .actor/actor.json + Dockerfile (apify/actor-* base) + src/main.*.
F56	Without CLI, fetch templates from apify/actor-templates via the raw manifest.json.
F48	Browser scenarios use apify/actor-node-playwright-chrome (or puppeteer variant), not plain actor-node.
F50	Smoke-test Standby Actors with curl, not apify call; set defaultTimeoutSecs.
F52	Default key-value store scoping is per-run; LIMITED_PERMISSIONS breaks cross-run state.

Why now

These are all failure modes with high observed occurrence across eval runs (multiple scenarios × multiple stacks each). The Skill overlay in the eval harness already carries a local version of these — this PR upstreams the general-audience subset so every Skill consumer benefits, not just the eval runner.

Test plan

• Render the SKILL.md locally and verify the new section appears where expected.
• Cross-check each teaching against FINDINGS.md in the eval repo (findings F19, F22, F33, F48, F50, F52, F56).
• Confirm no wording contradicts existing Best practices / Don't sections.