timeout failures in tc_init break systest logging and error reporting

[hyperthunk]

We're stuck somewhere between common_test and test_server here, but it's not easy to actually generate a decent error report when this happens. How best to do so isn't 100% clear to me right now, but my money is on a solution around issue #50 wherein the lifecycle of each participating component is governed in a unified way (e.g., an fsm based driver), because this way we will always know in what state every component has got itself into when any kind of failure occurs and we can drive the error reporting from this 'known state of the world' much more easily and robustly that way.

[hyperthunk]

In addition to the lacklustre information the ct logs give about such errors, somehow the ct_hook return value seems to be getting screwed up such that we see ct related noise in the console (and html logs) which shouldn't really be there:

[system]  ERROR: unable to start smoke_test

- - - - - - - - - - - - - - - - - - - - - - - - - -
Error detected: {system_under_test,start,{error,...}}
- - - - - - - - - - - - - - - - - - - - - - - - - -

[system]  smoke_test failed! Unexpected process exits detected:
[system]  smoke_test: {bad_return_value,
                          {hook_failed,
                              {local,rabbit_ha_test_utils,start_rabbit,[]},
                              {test_case_failed,
                                  "Unable to start rabbit on rabbit1@iske: {error,
{cannot_start_application,rabbit,
 {bad_return,
{{rabbit,start,[normal,[]]},
{'EXIT',
{rabbit,failure_during_boot}}}}}}\nPlease check the node's logs for more information."}}}
[system]  test case smoke_test failed.

The {bad_return_value, _} tuple seems to indicate that the systest ct_hook is not behaving properly, and whilst providing the on_start error info is useful, this is not the right format and the bad_return_value error tuple is actually a red herring and adds nothing but confusion for the test author.