Add context manager for invoke_later and invoke_when for gui testing

I have learned, from a separate project that, three main functions are very useful for GUI testing.

(1) A context manager that schedules a function to be invoked later. The invocation will be canceled when leaving the context

(2) A context manager that schedules a function to be invoked when a condition is fulfilled. The invocation will be canceled when leaving the context

(3) A function that starts the event loop and runs until a condition is fulfilled, with a timeout.

(3) is provided by `EventLoopHelper.event_loop_until_condition` (for Qt only?). (1) and (2) are not provided by pyface as context manager, I had to write my own.

Then in tests, I do this pattern a lot:
```
with call_later(wait=0.1, function=run_something_that_makes_condition_true), \
call_when(condition=condition, function=close_and_exit):
loop_until(condition=lambda: False, timeout=20.0)
```

The main idea about implementing these context managers is that they prevent test interactions: An invocation scheduled in the event loop by one test will never be run in another test.