Hackily use addSkip to track flakiness in common_utils.py (#78292)
### Problem:
The current way we detect flakiness is by aggregating results at the end of a job, which has worked so far but is inefficient and potentially inaccurate. We have also been delegating a workflow step towards doing this analysis at the end of every job.
### Solution:
This PR uses unittest.TestResult's addSkip method, which adds a skipped test every time we detect something is flaky. This way, we no longer need to aggregate anything and we can easily scan through the test reports and filter for skipped tests with flaky = True. Not only is this much faster to query for, it rids us of needing to figure out janky aggregation logic.
### Test plan:
I simulated a flaky test locally (test_async_python) and observed that:
With overriding signal ON (so flaky test = green):
- Successes pass are reported just as they normally are with no skips. [override_signal_normal_success.txt](https://github.com/pytorch/pytorch/files/8774012/override_signal_normal_success.txt)
- Failures fail and are reported as they are with no skips. [override_signal_all_fails.txt](https://github.com/pytorch/pytorch/files/8774010/override_signal_all_fails.txt)
- Flaky tests have expected failures + a success + a skip denoting the correct information. [override_signal_1_1.txt](https://github.com/pytorch/pytorch/files/8774005/override_signal_1_1.txt)
and [override_signal_2_1.txt](https://github.com/pytorch/pytorch/files/8774007/override_signal_2_1.txt)
With overriding signal OFF:
- Successes pass are reported just as they normally are with no skips. [report_only_one_success.txt](https://github.com/pytorch/pytorch/files/8774019/report_only_one_success.txt)
- Failures fail and are reported as they are with no skips. [report_only_all_fails.txt](https://github.com/pytorch/pytorch/files/8774018/report_only_all_fails.txt)
- Flaky tests have failures + unexpected successes + a skip denoting the correct
information. [report_only_3_1.txt](https://github.com/pytorch/pytorch/files/8774015/report_only_3_1.txt)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/78292
Approved by: https://github.com/suo