Add customizable behaviour for handling flaky tests

[artemnikitin]

Like #15, but for more granular control with different options. The potential list of options:

• If at least on 1 device tests are green, but all other devices are unavailable
• If at least on 1 device tests are green, but on some devices they are failing
• If at least on 1 device tests are green and don't care about anything else