feat: add idempotent ledger checkpoint system

[wheval]

📌 Summary

Adds a ledger checkpoint system to prevent duplicate event processing after indexer restarts.

🎯 Purpose / Motivation

When the indexer restarts mid-batch, it currently has no reliable mechanism to resume from the exact ledger and event offset. This causes duplicate trade records and gaps in trade history. A checkpoint system makes restart recovery deterministic and safe.

🛠️ Changes Made

• Add idempotent ledger checkpoint system to prevent duplicate event processing after indexer restart #632: Add idempotent ledger checkpoint system
• Added LedgerCheckpoint Prisma model with ledger, lastProcessedEventIndex, batchHash, completedAt
• Created checkpoint service with write, read, validate, and resume functions
• Non-blocking checkpoint writes with 5s timeout (falls back to logging on failure)
• Batch hash mismatch detection triggers warn log and full ledger reprocess
• Integrated checkpointing into event processor with ledger-grouped processing
• Added unit and integration tests for checkpoint operations

🧪 How to Test

1. Run pnpm test to verify all tests pass
2. Check that LedgerCheckpoint model is in Prisma schema
3. Verify checkpoint service writes/reads correctly via integration tests
4. Verify event processor groups events by ledger and checkpoints after each

⚠️ Breaking Changes

• None.

🔗 Related Issues

Closes #632

✅ Checklist

• Code builds successfully
• Tests added/updated
• No console errors
• Documentation updated (if needed)

[Chucks1093]

Nice work on this. The checkpoint system is a clean solution. Grouping events by ledger and writing the checkpoint after each batch makes restarts safe without any complex state to manage. The 5s timeout on checkpoint writes is a good call too, it keeps the main flow from stalling if the write is slow. Good test coverage as well.