[Feature] Optimize bulk load performance for CSV COPY inputs

[zhanglei1949]

Is your feature request related to a problem? Please describe.
Currently, bulk loading vertices/edges from CSV via COPY reads all batches into memory before insertion, and the bulk load process runs single-threaded. This leads to high memory usage and limits load throughput for large CSV datasets.

Describe the solution you'd like

• Fuse DataSource and BatchInsert so batches are consumed as they are produced, avoiding buffering the entire input in memory.
• Parallelize the bulk loading process across multiple threads to improve load throughput.

Describe alternatives you've considered
Keeping the current single-threaded, fully-buffered load path, which is simpler but does not scale for large datasets.

Additional context
Related to #734, which implements this optimization by refactoring csv_read_function, batch_insert_edge/batch_insert_vertex, loader_utils, and related storage/graph interfaces.

[github-actions]

🤖 Auto-triage: linked as sub-issue of #221 — Improve & Extend COPY FROM: schema, typing, performance, options (confidence: high).

Reason: The issue is about optimizing COPY FROM performance (memory usage, batching, throughput) for CSV inputs, which falls squarely under #221's scope of COPY FROM performance including batching and memory caps.

Auto-applied: type → Feature; labels → executor, store

If this is wrong, please unlink and re-assign — your correction improves this bot.

[zhanglei1949]

[loader] Total COPY time: 6433.64s (31 statements)
[loader] Finished in 15598.61s