Speed up mock_constructor by removing gc.collect

[deathowl]

Summary:
mock_aiog
0:00:00.064367
from this calling gc.collect took:
0:00:00.053297 and it's synchronous

since default slow callback detection is 0.1 if someone uses multiple mock_constructor calls these gc.collect-s add up and exceed the threshold

Differential Revision: D38073077

[facebook-github-bot]

This pull request was exported from Phabricator. Differential Revision: D38073077

[coveralls]

Pull Request Test Coverage Report for Build 2718898253

• 0 of 0 changed or added relevant lines in 0 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.002%) to 94.116%

---

Totals
Change from base Build 2710955248:	-0.002%
Covered Lines:	2639
Relevant Lines:	2804

---

💛 - Coveralls

[fornellas]

IIRC this GC had to be there for consistency at the check. I belive removing it, will lead to inconsistent corner cases where it'll complain mock will fail, wrongly.

[deathowl]

It did not trigger any test faillures on the internal tests,but actually the false slow calllback detection becaause of GC blocking the event loop caused more faulty signals, so the change is reasonable.

[fornellas]

I don't think i'd trigger any consistent build issues. IIRC, without this GC, depending on how objects are handled up to that point, there MAY be lingering references the GC would clean up, that, without cleanup, it would then trigger false errors. So, it'd be the kind of unexpected inconsistent behaviour, rather than always failing.

I recommend considering alternatives, such as wrapping this GC run with a temporary disable of the event loop block check.