Multi-frame tracing user interaction

[laike9m]

On the backend, frames are stored in a tree

UI interaction:

Picking frames

1. User clicks on a location in VSC
2. Extension sends a location to backend
3. Backend returns the first 5 frames that contain the specified code location, with
   the callsite location.
   (In the future we can extend the max number of candidates.)
4. User picks a frame
5. The identity of the picked frame is sent to backend
6. Backend sends back tracing results for the picked frame

Steps 2 ~ 5 won't happen if there's only one frame.

If there's no frame that matches the current location, nothing will happen.

Map tracing result to code
TBD. But for now, we should prevent frame selection process from happening (again) if
tracing is present and code location didn't go out of the frame's scope.

By default, we should show previous frame + current frame + 1-level frames derived
from the current frame. We will let users configure this on extension UI.

[laike9m]

Related #73