tests(bigquery): implement robust wait loop for socket leak tests

[chalmerlowe]

Intent

This PR addresses flakiness in the google-cloud-bigquery system tests that verify system resource cleanup (specifically open sockets).

The Problem

Some system tests in google-cloud-bigquery were performing instantaneous assertions on the number of open sockets immediately after client closure or garbage collection. These assertions are fragile because underlying transport layers (like gRPC) and OS socket reclamation can occur asynchronously in background threads. This leads to intermittent failures (flakiness) especially under varied environmental loads or parallel execution.

Solution

Refactored the fragile instantaneous checks in test_client.py and test_magics.py to use a Robust Wait-Loop pattern.

• The tests now wait up to 3 seconds, checking periodically, to allow asynchronous cleanup to complete before asserting.
• This pattern is already proven effective in other packages like google-cloud-pubsub.
• Added explicit gc.collect() to test_magics.py to align with best practices for resource verification tests.

Note

Also includes some linting updates. Sorry.

[gemini-code-assist]

Warning

Gemini encountered an error creating the review. You can try again by commenting /gemini review.

[parthea]

/gemini review

[gemini-code-assist]

Code Review

This pull request reorganizes imports and introduces retry loops in several socket leak tests to wait up to 3 seconds for connections to close before making assertions. The feedback suggests moving the gc.collect() call inside these retry loops to ensure any objects that become eligible for garbage collection during the sleep intervals are collected and their sockets are properly released.

[gemini-code-assist]

To ensure that any socket-holding objects that become eligible for garbage collection during the sleep intervals are actually collected and their sockets released, gc.collect() should be called inside the wait loop rather than only once before it.

Suggested change

	for _ in range(30): # Wait up to 3 seconds
	conn_end = current_process.net_connections()
	conn_count_end = len(conn_end)
	if conn_count_end <= conn_count_start:
	break
	time.sleep(0.1)
	for _ in range(30): # Wait up to 3 seconds
	gc.collect()
	conn_end = current_process.net_connections()
	conn_count_end = len(conn_end)
	if conn_count_end <= conn_count_start:
	break
	time.sleep(0.1)

[gemini-code-assist]

To ensure that any socket-holding objects that become eligible for garbage collection during the sleep intervals are actually collected and their sockets released, gc.collect() should be called inside the wait loop rather than only once before it.

Suggested change

	for _ in range(30): # Wait up to 3 seconds
	conn_end = current_process.net_connections()
	conn_count_end = len(conn_end)
	if conn_count_end <= conn_count_start:
	break
	time.sleep(0.1)
	for _ in range(30): # Wait up to 3 seconds
	gc.collect()
	conn_end = current_process.net_connections()
	conn_count_end = len(conn_end)
	if conn_count_end <= conn_count_start:
	break
	time.sleep(0.1)

[gemini-code-assist]

To ensure that any socket-holding objects that become eligible for garbage collection during the sleep intervals are actually collected and their sockets released, gc.collect() should be called inside the wait loop rather than only once before it.

Suggested change

	for _ in range(30): # Wait up to 3 seconds
	conn_end = current_process.net_connections()
	conn_count_end = len(conn_end)
	if conn_count_end <= conn_count_start:
	break
	time.sleep(0.1)
	for _ in range(30): # Wait up to 3 seconds
	gc.collect()
	conn_end = current_process.net_connections()
	conn_count_end = len(conn_end)
	if conn_count_end <= conn_count_start:
	break
	time.sleep(0.1)