Add native asyncio cursor support (Phase 1)

[laughingman7743]

Summary

• Add AioCursor and AioDictCursor that use native asyncio instead of ThreadPoolExecutor
• Uses asyncio.sleep() for polling and asyncio.to_thread() for boto3 calls, keeping the event loop free
• Add pyathena.aconnect() async entry point and AioConnection with async context manager
• Add AioResultSet with async create() factory pattern (cannot await in __init__)
• 16 integration tests passing

Motivation

Closes #662 (Phase 1). The existing AsyncCursor uses ThreadPoolExecutor which blocks threads during time.sleep() polling. Native asyncio support allows proper integration with async frameworks (FastAPI, aiohttp, etc.).

Scope (Phase 1 only)

This PR covers the core cursor types only:

• AioCursor / AioDictCursor
• AioResultSet / AioDictResultSet
• AioBaseCursor / AioConnection

Deferred to Phase 2: Specialized cursors (pandas, arrow, polars, s3fs, spark) and metadata operations (list_databases, list_table_metadata, etc.).

Key design decisions

Decision	Choice	Rationale
Blocking calls	asyncio.to_thread(retry_api_call, ...)	Zero new deps, works with any boto3 version
Result set init	async create() classmethod	Cannot await in __init__
Naming	Aio prefix (AioCursor)	Async prefix already taken by thread-based cursor
Cancellation	asyncio.CancelledError	Correct async idiom (replaces KeyboardInterrupt)
Inheritance	Extend sync BaseCursor	Reuse _build_* methods, __init__, constants

Usage

import pyathena

conn = await pyathena.aconnect(
    s3_staging_dir="s3://bucket/path/",
    region_name="us-east-1",
)
async with conn.cursor() as cursor:
    await cursor.execute("SELECT * FROM my_table")
    rows = await cursor.fetchall()
    async for row in cursor:
        print(row)

Test plan

• make fmt passes
• make chk (ruff + mypy) passes
• 16 async integration tests passing (tests/pyathena/aio/test_cursor.py)
• CI pipeline passes

🤖 Generated with Claude Code