Record the buffer that the command ran in and display

[lastquestion]

In #46 , we dropped the command stack display code and started showing only the command that ran. This buckets all commands in all buffers and all situations into one. Often times things are slow only in one buffer, etc.

1. Record the buffer that the action came from
2. Expand explain-pause-top to support trees and allow expansion of nodes.

Note that this basically implements classic "bottom up" call tree view of a profiler view.
Obvious expansions then are reversed, so "top down" from a buffer, process, etc.

While doing this, clean up the command stringification logic as well.
Before attempting this, implement unit tests for the display code.