Dependency is not considered in re-run tests

With Plainbox 0.21, we now have the possibility to re-run failed tests at the end of a session (before submitting it).

However, these re-run tests currently do not consider test case dependencies, which causes some problems.

For example:
A tester failed the following cases and is prompted to re-run these cases before submitting report:
- keys/wireless
- suspend/keys_after_suspend/wireless (this case depends on suspend/suspend_advanced)

Expected re-run scenario:
(1) [If possible] Plainbox triggers system to auto reboot to get clean "before s3" state
(2) keys/wireless is run
(3) suspend/suspend_advanced is run (so that cases which depend on it (i.e. after-suspend cases) do run in an after-suspend environment)
(4) suspend/keys_after_suspend/wireless is run

Actual re-run scenario:
(1) keys/wireless is run
(2) suspend/suspend_advanced is *not* run because it previously passed
(3) suspend/keys_after_suspend/wireless is run directly after keys/wireless