Refactor test code structure

Currently, the test code layout is messy, with so many convoluted (and possibly circular) dependency directions for code execution. We should spend some time to clean up the codebase to make it easier to follow. My biggest concern is around the `metric_value_benchmark` test directory and everything in there. This is the biggest headache at the moment.

Start by inspecting the metric_value_benchmark test directory and tracing the execution dependencies described in the issue. Map the current test-code layout before changing it. The work is done when the convoluted or circular dependency directions are cleaned up and the test structure is easier to follow.

Written by the indexing model from the issue text.