fix: streaming database dump with pagination for large databases

[aadityaranjan01]

/claim #59

Fix: Large database dumps fail due to memory and timeout limits

This PR replaces the current in-memory dump implementation with a streaming-based approach to support large databases reliably.

Problem

The existing /export/dump implementation:

• loads entire tables into memory (executeOperation → cursor.toArray())
• builds a full dump string before responding
• fails for large databases due to memory limits and timeouts

Solution

This PR introduces a minimal, streaming-based fix:

• Uses ReadableStream to stream dump output progressively
• Fetches rows in batches using LIMIT/OFFSET
• Adds ORDER BY rowid to ensure stable pagination
• Includes breathing intervals to prevent Durable Object blocking
• Keeps all changes scoped to src/export/dump.ts

Key Benefits

• Works for arbitrarily large databases
• Avoids memory overflow
• Avoids timeout failures
• No new infrastructure required
• Minimal diff, no changes to shared utilities

Example Usage

curl --location 'https://<your-worker>/export/dump' \
--header 'Authorization: Bearer <token>' \
--output database_dump.sql

Notes

• Does not modify executeOperation or other export routes (JSON/CSV remain unchanged)
• Designed as a safe, incremental improvement without introducing async jobs or R2 dependencies