fix: Address code review issues in select-algorithm samples

[diberry]

Summary

Comprehensive code quality fixes across all 5 select-algorithm samples (Python, TypeScript, .NET, Go, Java) based on 8-agent automated review cycle. This PR passed 7 review rounds before achieving unanimous approval from all 8 review agents.

Changes by Language

All Languages

• Replace .env files with shell �xport commands (no dotenv dependencies)
• Fix data file paths to use local data/Hotels_Vector.json (relative to sample root)
• Document LOAD_SIZE_BATCH env var (default: 100) consistently
• Remove connection string references — all samples use passwordless/Entra ID
• Fix README structure: consistent sections across all languages

.NET (select-algorithm-dotnet)

• Rename config section MongoDB → DocumentDB in appsettings.json
• Read EmbeddingDimensions from config (was hardcoded)
• Remove phantom EmbeddingSizeBatch config reference
• Document DocumentDB__* env var override pattern (double underscore)

Java (select-algorithm-java)

• Read LOAD_SIZE_BATCH from environment (was hardcoded to 100)
• Fix
  etryWrites=false in connection setup

TypeScript (select-algorithm-typescript)

• Remove �nv:init script from package.json
• Fix data path to local data/ (not ../data/)

Python (select-algorithm-python)

• Fix README to run from sample root (python src/compare_all.py)
• Data path resolves relative to root, not src/

Go (select-algorithm-go)

• Replace .env references with shell exports
• Consistent data path documentation

Review Process

This PR went through a 3-phase, 8-agent automated review cycle:

Phase	Focus	Rounds	Duration
Phase 1	Code correctness	7 rounds	~1h 13m
Phase 2	Content alignment (PR #240)	3 rounds	~18m
Phase 3	Cross-consistency (code ↔ content)	4 rounds	~12m

Total: 14 review rounds, 9 fix commits, ~2h 13m wall-clock

Review Agents (all 8 approved)

1. PR Reviewer — structural quality, PR hygiene
2. Tech Accuracy — factual correctness vs. DocumentDB docs
3. Copy Editor — language clarity, Microsoft style
4. Quickstart Reviewer — developer experience, runability
5. Fact Checker — cross-reference upstream sources
6. Cross-Language Consistency — parity across 5 languages
7. Content PR Reviewer — Learn platform compliance
8. AI Prompts Engineering — prompt/schema validation

Final verdict: ✅ ALL 8 APPROVE

Related PRs

• Content: MicrosoftDocs/nosql-docs-pr#240 (article fixes, also 8-agent approved)
• Skills: diberry/project-dina#407 (skill definition fixes)

Conventions Enforced

• DOCUMENTDB_* prefix for all env vars (never MONGO_*)
• Passwordless auth via DefaultAzureCredential (never connection strings)
• Shell exports only (no .env files, no dotenv libraries)

etryWrites=false across all languages

• Cleanup via VS Code extension (never mongosh)