Important changes:

* Mocha 8 and later are now supported. Previously, it's pending tests
  were incorrectly reported as undefined or failing.

* `QUnit.todo()` is now supported. Previously, these intentional
  failures were reported by browserstack-runner as failing tests.

  With js-reported 2.1 the runEnd report counts them as `todo` instead
  of as `failed`.

The API that browserstack-runner depends on has not changed since 1.1.0.
<https://github.com/js-reporters/js-reporters/blob/v2.1.0/CHANGELOG.md>.

Specifically, browserstack-runner captures the following data in
lib/_patch/reporter.js, and proccesses it in lib/server.js under
the "_progress" and "_report" handlers:

* The `testEnd` event, reading:
  - `name` string,
  - `suiteName` string,
  - `status` string,
  - `errors` array.

* The `runEnd` event, reading:
  - `status` string,
  - `testCounts` object with ints `total`, `passed`, `failed`, `skipped`,
  - `runtime` int.

These still exist the same way in the latest release:
<https://github.com/js-reporters/js-reporters/blob/v2.1.0/spec/cri-draft.adoc#runend>

Fixes #248.
Ref #247.

The update to js-reporters 2.1.0 in the previous commit has fixed
the most important problem, which is the build status reported
by the `runEnd` event, handled via "_report" submission.

However, the todo tests were still printed as errors in the output
which is confusing.

Before:

```
[Windows 8, Chrome 91.0] Error: "test.todo.each [0]" failed
  Expected: true
  Actual: false
[Windows 8, Chrome 91.0] Passed: 332 tests, 321 passed, 0 failed, 7 skipped; ran for 1255ms
```

After:

```
[Windows 10, Firefox 88.0] Passed: 332 tests, 321 passed, 0 failed, 7 skipped; ran for 1910ms
```

Fixes #247.

Ref #263.