Small reorganization and fixes for test_spill

[pentschev]

Do a minor reorganization on how client.run() is invoked and the centralize functions to check host and disk chunks.

The failures seem related to del not cleaning up objects in time, thus invoke garbage collection after del until it memory is actually released or the test times out. Local tests seem that invoking garbage collection once or twice is enough to prevent the test from failing.

[madsbk]

Do we want a max number of tries?

Suggested change

	while True:
	try:
	await client.run(
	delayed_worker_assert,
	0,
	0,
	0,
	)
	except AssertionError:
	gc.collect()
	else:
	break
	for _ in range(10):
	try:
	await client.run(
	delayed_worker_assert,
	0,
	0,
	0,
	)
	except AssertionError:
	gc.collect()
	else:
	return

[pentschev]

I thought about this, but then I decided in favor of a timeout, as per description: "thus invoke garbage collection after del until it memory is actually released or the test times out.".

I think timing out is overall more appropriate than an arbitrarily-defined count. What do you think?

[pentschev]

The timeout is defined by the gen_test decorator, and that reminded me we can now set it back to the 30 seconds it was before.

[madsbk]

Sounds good

[madsbk]

LGTM, thanks @pentschev

[pentschev]

Thanks @madsbk !

[pentschev]

/merge