latest

Consider worker failures as cancelled executions in tests

Summary:
When a sandcastle job that runs tests using a worker tool is cancelled, the sandcastle worker will kill buck, which in turn will kill the worker, and once that is killed, it'll kill the test runner too. In the time between the worker is killed and the test runner is killed, the test runner still attempts to run tests via the worker, but the worker returns `GatherOutputStatus::SpawnFailed` (see https://www.internalfb.com/code/fbsource/[88a84f351975921f560270e1096a8687d6f62160]/fbcode/buck2/app/buck2_execute_impl/src/executors/worker.rs?lines=479%2C492). This causes the test runner to report a ton of fatal test results (e.g. https://www.internalfb.com/intern/testinfra/diagnostics/12666374014604584.844425084763262.1735210981).

To avoid this, I'd like to tell the test runner that test execution was cancelled instead of completed (un)successfully. This will produce omitted test results.

Reviewed By: shashkambham

Differential Revision: D67734577

fbshipit-source-id: fc7616ec2cf26260a46ff721474e52e63da254da