VolumeTracker should deal with Phi nodes appropriately

[tysolov]

The volume tracker does not appropriately handle phi nodes.

The issue arises when a phi node exists in a loop header, where there is a use prior to a def, and has the following structure:

before_loop:
    a0 = ...
loop_header:
    a1 = phi(a0, a1)
loop_body:
    ... = use a1

A test case has been added to the volume tracking testing harness using tests/assays/probabilistic_pcr.bs as the model.