pytest timeout is not respected when test is executing long-running SQL query

[alexanderlaw]

This issue is a spin-off of #3760.

A test as simple as:

    env = neon_env_builder.init_start()
    endpoint = env.endpoints.create_start("main")

    conn = endpoint.connect()
    cur = conn.cursor()
    cur.execute("SET statement_timeout='600s'; SELECT pg_sleep(400);")

doesn't get interrupted timely, with the default pytest timeout (300):

...
2025-01-09 14:18:22.756 INFO [neon_fixtures.py:4167] neon.file_cache_size_limit = 1MB
------------------------------------------------------- Captured log teardown -------------------------------------------------------
2025-01-09 14:25:02.886 INFO [neon_fixtures.py:886] Cleaning up all storage and compute nodes
...
======================================================
FAILED test_runner/regress/test_timeout.py::test_timeout[debug-pg17] - Failed: Timeout >300.0s

A similar (worse) situation is possible when a backend just hangs and doesn't process interrupts for a long/indefinite time.