[Feature Request]: Provide asynchronous file relations and Arrow analytics APIs #30066

Existing issue check

• Checked relevant Python SDK issues; this report concerns the SDK client path.

Environment

Repository main 3e50e1b32d8da61bbbf70c89989fb08b671809de, Python 3.12, SQLAlchemy 1.4/2.x. Initial findings were reproduced offline with synthetic inputs; real-service validation will be linked from the implementation PR.

Priority: P1. Related: #30049 and draft implementation #30054 (synchronous scope).

Problem and requirements

AsyncClient has no equivalent of the proposed synchronous analytics interface. Applications should not need blocking file/result operations on the event loop.

Provide async context-managed file relations, async Arrow batch iteration, DataFrame/Arrow upload and import with equivalent type/precision/commit-outcome contracts. Honor cancellation/deadlines and pool backpressure. Preserve optional dependencies and explicit shared-path limitations. No Flight or zero-copy promise is implied.

Design considerations

Use SQLAlchemy async streaming/aiomysql unbuffered retrieval; bound Arrow conversion and staging work and move blocking I/O off the event loop where appropriate. Retain transaction settings through COMMIT, distinguish unknown outcomes and define cancellation cleanup. Do not merely wrap complete synchronous result materialization in a thread.

Acceptance tests

Compare complete values and schemas with synchronous APIs; measure event-loop heartbeat during large transfers; cancel during staging/LOAD/read/COMMIT, exhaust a small pool, inject network stalls and verify outcome/resource ownership. Run dependency, same-host dual-CN and permissions matrices with no unexpected skips.

Implementation tracking

Design/implementation follow-up remains open. This requirement is not claimed as implemented by the current bug-fix/performance PRs. Related synchronous analytics draft: #30054.

Async connection pool configuration

Expose pool_size, max_overflow, pool_timeout and pool_recycle rather than retaining fixed testing defaults. Verify bounded backpressure and cancellation with a one-connection pool; preserve documented default behavior.