docs: add requests table S3 archival design doc (#175)

[Achyuth0506]

Design document for Issue #175 - periodic archival of the request table to AWS S3.

What's included

• Problem statement and goals covering three consumers - internal analytics, RAG agent, and third-party vendors
• Three-stage pipeline architecture - Collection, Scrubbing, and Serving with separate S3 staging areas
• Schema mapping from actual DDL (virginia_dev_saayam_rdbms.request) to Parquet with logical-to-real column translation table
• PII handling with explicit per-column decisions based on real schema - HMAC hashing, city/state coarsening, boolean derivations for doc and audio fields
• Dimension/lookup table resolution strategy for six FK columns
• Four-layer free-text scrubbing approach with honest residual risk acknowledgment
• Vendor serving design with three access model options and rationale for Option C
• Four alternatives considered with tradeoffs
• Cost estimate at current and 10x scale with per-service breakdown
• Six-phase rollout plan with rollback strategy
• 18 open questions tracked with owners and status

Pending Inputs

• Row count and growth rate
• Read replica availability
• DB version (RDS vs Aurora, major version)
• last_update_date UPDATE trigger confirmation
• to_public column confirmation

Acceptance Criteria Checklist

• Readable by someone outside data eng
• Non-obvious tradeoffs surfaced
• PII handling has explicit per-column decisions
• Cost estimate at current and 10x scale included
• At least two alternatives documented with rationale
• Implementation plan concrete enough to estimate in story points
• Vendor access model confirmed at design review
• Design reviewed and signed off by data eng team