Tweak get_stack_trace() API

[living180]

Replace the depth argument to get_stack_trace() with a more intuitive skip argument. Update the cache and SQL panels to supply a skip argument to get_stack_trace() to ensure that the stack traces are clean even when HIDE_IN_STACKTRACES is empty.

[matthiask]

I'm not completely sure if skip is clearer than depth. Maybe skip_stack_levels but in this case we could copy the stacklevel name from warnings.warn?

[living180]

How about skip_frames? I don't want to copy stacklevel because my skip argument has slightly different semantics. stacklevel from warnings.warn() and depth from sys._getframe() both default to 1, and the reference point is from the frame of warnings.warn() or sys._getframe(). But the caller doesn't care about the frames of the functions it is calling to get the trace. It wants either a trace starting with itself, or a lower frame if it (or possibly one or more of its callers) shouldn't be displayed to the user. That's why I switched to a default value of 0 for skip. When I first added the depth argument to the get_stack_trace() calls for the cache and SQL panels, it was depth=3, which felt confusing and I really had to think about it, and that was right when I was writing the code (not coming back and trying to understand it weeks or months later). Now with skip=2, the meaning seems clear to me: skip the current frame (which calls get_stack_trace()) plus one more frame (the caller of the current frame).

Does this make sense?

[matthiask]

Yes, makes sense. Thanks for the explanation!

I think the change is fine as-is.