Pin down common way to skip tests based on toolkit

Ref: #867

Currently we skip many tests so that they only run on a particular toolkit. However, this is done in a variety of ways and varies from test to test.

(Copied from 867 PR description: )
Looking into this also made me realize that there is no consistent way this type of thing is done across the codebase. For example:
https://github.com/enthought/pyface/blob/8cc824ec59f68ae6ebe6a8dcfea006d62a66811f/pyface/ui/wx/grid/tests/test_simple_grid_model.py#L12-L20
https://github.com/enthought/pyface/blob/8cc824ec59f68ae6ebe6a8dcfea006d62a66811f/pyface/action/tests/test_traitsui_widget_action.py#L25
https://github.com/enthought/pyface/blob/8cc824ec59f68ae6ebe6a8dcfea006d62a66811f/pyface/tests/test_api.py#L17-L18
https://github.com/enthought/pyface/blob/8cc824ec59f68ae6ebe6a8dcfea006d62a66811f/pyface/tasks/tests/test_dock_pane_toggle_group.py#L26

sometimes trying the import, sometimes working with `toolkit` from `pyface.toolkit` sometimes through `ETSConfig.toolkit`, etc. It is pretty inconsequential (they all do what we want), but it would be good to pin down a consistent way to skip tests based on toolkit.