Allow sending "mouse move" with left/right down with EnableTestAssistant

[aaronayres35]

fixes #460

This PR adds left_down and right_down kwargs to mouse_move so that the test assistant can simulate moving the mouse with either the left or right button down. As mentioned on the issue, there are cases where we expect mouse_move events to have {left/right}_down set to True. This PR also adds a simple test to check that the traits are set if we call mousee_move with the new kwargs. Additionally, the press_move_release method is updated so that left_down is True for the mouse_move generated event itself.

[aaronayres35]

It seems in some cases mouse move events are not being handled. Need to investigate further.

[jwiggins]

You just need to add left_down=True to the mouse_move calls in the failing tests.

That said, it would be nice to know why this broke those tests to begin with.

[jwiggins]

We should maybe add support for middle_down while we're here, no?

[aaronayres35]

Previously, if {left/right}_down was not set, they defaulted to Undefined. Thus, in this code:

enable/enable/tools/drag_tool.py

Lines 155 to 157 in 2160808

	button_down = getattr(event, self.drag_button + "_down")
	if state == "nondrag":
	if (button_down

button_down would evaluate to True, event if {left/right}_down was never set. Now, this PR explicitly sets them to False, hence the new test failures.

Unfortunately that suggests this PR may break existing code out in the wild that unknowingly depended on this buggy behavior

Perhaps a less disruptive fix would be to have the left_down/ right_down kwargs default to Undefined like they used to, instead of explicitly making them False? The dependence on Undefined being True surely seems like a bug, but we may not want to break code. @jwiggins what do you think?

[jwiggins]

@jwiggins what do you think?

Hard question. I don't like the bug exposed here but we might want to keep the default values as Undefined so that we don't break downstream CI for dependent projects.

[aaronayres35]

@jwiggins what do you think?

Hard question. I don't like the bug exposed here but we might want to keep the default values as Undefined so that we don't break downstream CI for dependent projects.

That makes sense, I will make that change and then open a separate issue for the use of Undefined

[jwiggins]

LGTM