feat: implement true batch writing for parquet writer

[Sanjay-Kirti]

Description

This PR implements true batch writing optimization for the Parquet writer, addressing performance issues identified in the current implementation. Instead of processing records individually in a loop, the implementation now groups records by partition and writes entire batches in single operations.

Problem Solved:
The existing implementation was calling the parquet writer once per record, which doesn't leverage the underlying parquet-go library's batch capabilities, leading to:

• Inefficient I/O operations due to single-record writes
• Increased overhead and reduced throughput for large datasets
• Suboptimal compression ratios
• Poor utilization of parquet format's performance features

Solution Implemented:

• Partition-wise batch grouping: Records are collected and grouped by partition path
• Single timestamp per batch: Efficient timestamping with one time.Now() call for entire batch
• True batch writes: Each partition gets ONE write operation instead of N individual writes
• Backward compatibility: Same function signature and error handling patterns

Performance Results:

• ✅ 1000 records processed in 2.6562ms (verified by performance tests)
• ✅ Dramatically reduced I/O operations: From N writes to P(partitions) writes
• ✅ All records in batch share same timestamp, proving true batching behavior

Fixes #477

Type of change

• New feature (non-breaking change which adds functionality)

How Has This Been Tested?

Comprehensive test suite created and all tests pass:

• Basic batch writing test: Multiple records batched correctly with shared timestamps
• Empty batch handling: Edge case properly handled without errors
• Single record processing: Works correctly with single-record batches
• Performance test: 1000 records processed in 2.6562ms, demonstrating significant improvement
• Partition file creation: Verifies correct partition file management
• Compilation test: go build ./destination/parquet/... passes
• Full test suite: go test ./destination/parquet/... -v passes with all scenarios

Related PR's (If Any):
This PR supersedes the previous attempt #529 which had GitHub conflict resolution issues. This is a clean implementation from scratch with:

Proper batch writing logic (not just API syntax)
Comprehensive test coverage
Proven performance improvement

---

Note

Replaces per-record writes with partition-wise batch writes using a single batch timestamp and grouped GenericWriter calls.

• Parquet writer (destination/parquet/parquet.go):
  • Write:
    • Batch-process records: assign one OlakeTimestamp per batch and group by partition via getPartitionedFilePath.
    • Create partition files on-demand; write each partition's batch in a single GenericWriter call (any vs types.RawRecord).
    • Early return on empty input; updated error messages for batch operations.

^{Written by Cursor Bugbot for commit 6e7109d. This will update automatically on new commits. Configure here.}

[Sanjay-Kirti]

I humbly apologise for creating a chaos.
@shubham19may Let me know your reviews.
@hash-data

[Sanjay-Kirti]

@hash-data Is this MongoDB/Iceberg test failure also happening on the main/staging branch? The error suggests it's a test cleanup issue unrelated to the parquet batch writing changes.