feat(pipeline): hardened v5 pipeline with staging+MERGE

[thalescb]

What

Production-grade onchain analytics ingestion pipeline replacing the prototype v1.

Changes

Pipeline v5 (new, pipeline-v5/):

• Staging table + MERGE write path -- idempotent, atomic, free (replaces streaming inserts)
• Block-aligned chunking -- no data loss on mid-block crash
• IngestionStatus + PipelineRuns BigQuery observability tables
• Slack alerting on failure and data staleness (>36h threshold)
• Structured JSON logging with run_id tracing
• Retry with exponential backoff on HyperSync and BQ transient errors
• Two CLI modes: daily (append to chain tip) and backfill (with --from/--to)
• Structured exit codes: 0=success, 1=partial, 2=failure
• Contract config registry -- adding a contract = adding a config entry

Phase 0 fixes (pipeline/index.ts):

• Inter-batch delay (200ms) preventing HyperSync 429 rate-limiting
• Global progress tracking surviving throws
• Exponential backoff with jitter on BQ insert retry
• Structured exit codes

Scheduling (.github/workflows/pipeline-daily.yml):

• GitHub Actions cron at 01:00 UTC daily
• Concurrency group prevents overlap
• dbt refresh step ready (commented, enable when ready)

Validation

• Daily mode: both contracts process, exit 0, freshness alerts fire correctly
• Backfill: 54 events fetched/staged/merged successfully
• Idempotency: re-running same range produces 0 new rows (MERGE dedup works)
• PipelineRuns table populated with correct run metadata
• Slack webhook tested and delivering alerts

Setup

Required GitHub Secrets for the cron workflow:

• ENVIO_API_TOKEN -- HyperSync API token
• GCP_SA_KEY -- GCP service account JSON key
• SLACK_WEBHOOK_URL -- Slack incoming webhook URL

Relates to #40