CI: PowerShell startup intermittently fails with `procargs` errno 5 on macOS

### Android framework version

N/A — this is an Azure Pipelines agent/PowerShell startup failure that occurs before the test-slicing script runs.

### Affected platform version

Azure Pipelines macOS agent in the `MSBuild Emulator Tests macOS > MSBuild+Emulator 7` lane.

Observed in [public build 1579507](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1579507&view=results), task `Slice unit tests with filter`.

### Description

The `PowerShell@2` task failed while starting PowerShell, before executing the test-slicing script. The startup path reported a failed macOS `procargs` call (`errno 5`) and threw `Microsoft.PowerShell.ManagedPSEntry+StartupException`.

This appears to be an intermittent hosted-agent/process-startup failure rather than a test failure. It made the `MSBuild+Emulator 7` CI lane fail without running its assigned tests.

### Steps to Reproduce

No deterministic local reproduction is currently known.

1. Run the public `dotnet-android` pipeline on the macOS hosted agents.
2. Observe the `Slice unit tests with filter` PowerShell task in an MSBuild emulator lane.
3. In the affected run, PowerShell fails during startup before the script executes.

### Did you find any workaround?

Rerun the affected CI job or pipeline.

### Relevant log output

```text
Call to 'procargs' failed with errno 5
Microsoft.PowerShell.ManagedPSEntry+StartupException
PowerShell exited with code 'null'.
```

The failure occurred in [build 1579507](https://dev.azure.com/dnceng-public/public/_build/results?buildId=1579507&view=results), `MSBuild Emulator Tests macOS > MSBuild+Emulator 7 > Slice unit tests with filter`.

Start with the Azure Pipelines macOS lane, specifically the `MSBuild+Emulator 7` `Slice unit tests with filter` PowerShell@2 task and public build 1579507 logs. Investigate the intermittent `procargs` errno 5 startup failure and define a reliable mitigation so the task starts and assigned tests run instead of requiring a rerun.

Written by the indexing model from the issue text.