Path successor result wrong when a component needs to be popped.

[sgwilym]

I think this may have happened when Paths were refactored to use HeapEncoding.

Given the following path:

 PathRc([PathComponentBox([8]), PathComponentBox([0, 255]), PathComponentBox([])])

the successor is computed to be:

PathRc([PathComponentBox([8]), PathComponentBox([1, 0])])

This is wrong. The successor should be:

 PathRc([PathComponentBox([8]), PathComponentBox([1])])

The culprit is somewhere in this block.