No cleanup when the test process is killed by a fastFail quality gate on Windows during `allure run`

**Describe the bug**

Given the following sequence of actions:

1. We are on Windows.
2. We define a `{ maxFailures: x, fastFail: true }` quality gate in the Allure configuration file.
3. We invoke `allure run`.
4. The quality gate rule is triggered.

Then the test process is terminated without having a chance to perform a cleanup. For example, `playwright test` is terminated with no `afterEach`/`afterAll` functions executed.

This may have the following consequences:

1. No proper cleanup (db reset, tmp files removed, etc).
2. No Allure metadata from `afterEach`/`afterAll` functions.

**To Reproduce**

Steps to reproduce the behavior (on a Windows machine):

1. Clone [the repo](https://github.com/delatrie/allure3-409).
2. Run `npm install`.
3. Run `npm allure-run`.

Messages from `afterEach` and `afterAll` are both missing in the standard output.

Additionally, compare the results from Linux, MacOS, and Windows here. The Windows test result is missing the stdout attachment.

**Expected behavior**

There are two lines in the standard output:

```
afterEach
afterAll
```

**Screenshots**

On Linux (`afterEach` and `afterAll` messages are here):

On Windows (no `afterEach` and `afterAll` messages):

Later.

**Additional context**

Later.

Reproduce the issue on Windows with `npm install` and `npm allure-run`, then compare the output with Linux or macOS. Start from the `allure run` entry point and the fastFail quality-gate path; done means `playwright test` runs both `afterEach` and `afterAll`, and the Windows result retains the stdout attachment.

Written by the indexing model from the issue text.