feat: Add batched processing for SQLite converter to prevent OOM

[taariq]

Summary

• Adds memory-efficient batched processing for SQLite to PostgreSQL migration
• Uses BatchedTableReader with rowid-based pagination to read tables in chunks
• Automatically calculates optimal batch size based on available system memory
• Prevents OOM errors when migrating large SQLite databases (7M+ rows)

Problem

The SQLite converter previously loaded all rows into memory before inserting to PostgreSQL. For large tables (e.g., 7.2M rows in a 2.3GB file), this required 8-16GB of RAM, making the tool unusable on typical machines.

Solution

Now processes rows in batches:

1. Read batch from SQLite (default: auto-calculated based on memory)
2. Convert batch to JSONB format
3. Insert batch to PostgreSQL
4. Repeat until all rows processed

Memory usage stays constant regardless of table size.

Changes

• src/sqlite/reader.rs: Add BatchedTableReader struct and read_table_batch() function
• src/sqlite/converter.rs: Add convert_table_batched() async function
• src/commands/init.rs: Update init_sqlite_to_postgres() to use batched processing
• Add comprehensive unit tests for batched reader

Testing

• ✅ All unit tests pass (14 new tests for batched reader)
• ✅ Clippy clean
• ✅ Formatter clean

Fixes #77