feat(cognee): add Cognee + Valkey cookbook and sample code

[atao2004]

Summary

Migrates the archived Cognee + Valkey cookbook (PR #15, archive/main-pre-reorg branch) to the current CONTRIBUTING.md cookbook structure.

Upstream integration: topoteretes/cognee-community#46 (merged, released on PyPI as cognee-community-vector-adapter-valkey==0.1.3).

Changes from the archived version

• Vendor neutrality: default LLM/embedding provider flipped from AWS Bedrock to Ollama (local, free, no account needed). OpenAI and Bedrock are now clearly labeled, collapsed <details> alternatives with AWS affiliation disclosed.
• Pinned valkey/valkey-bundle:latest → 9.1.0.
• 03-production.md reworked from an ElastiCache-only deployment guide into generic production patterns (TLS, HNSW tuning, monitoring, error handling), with both AWS (ElastiCache/MemoryDB) and Google Cloud (Memorystore for Valkey) named as examples alongside generic managed-provider guidance.
• Added sample/docker-compose.yml and a CI-friendly test suite (test_valkey_adapter.py) — neither existed in the archive. The test suite exercises the Valkey vector adapter directly with a deterministic stub embedder, so it needs no LLM or Ollama to run in CI.
• Added sample/common.py to share the Ollama/Cognee bootstrap config across the three sample scripts (previously would have been triplicated).
• Dropped meta.json — not part of the current cookbook structure.

Validation

• Full add → cognify → search pipeline run end-to-end against valkey-bundle:9.1.0 + local Ollama (qwen2.5:7b + nomic-embed-text) — smaller models (llama3.2:1b/3b) were tried first and found unreliable for Cognee's structured-output requirements; documented in the cookbook.
• test_valkey_adapter.py (6 tests) passing in a clean-room venv against requirements.txt exactly as written, plus the actual docker-compose.yml.
• markdownlint-cli2: 0 errors. ruff check: clean.
• Went through a full internal QA review pass (spec-compliance, DRY, security, anti-pattern, link-integrity) — findings fixed: missing ## Step N headings in 02, missing How It Works/Configuration Reference tables in 01, single-vendor cloud guidance in 03, duplicated config extracted to common.py, and an incorrect safe_search return type annotation.

Pre-Submit Checklist

• Builds and runs from a clean clone
• Runs against current stable Valkey (Valkey Bundle 9.1.0)
• No private links, internal references, or local paths
• No marketing language or uncited claims
• Vendor-neutral by default (affiliation disclosed where applicable)
• README explains the Valkey concept demonstrated, prerequisites, and how to run
• Dependencies pinned to specific versions
• CI validates the sample

Known follow-up (not addressed in this PR)

cookbook-python.yml's dorny/paths-filter filter list doesn't include the workflow file path itself, even though the top-level trigger does — meaning a PR that only edits this CI file wouldn't exercise the jobs it configures. Left as-is since this file is the documented temporary-copy pattern pending PR #45's shared Python workflow, which is expected to supersede it.

[jbrinkman]

Closing this draft PR as part of a cookbook program cleanup.