Set commute_level for workplace assignment distinctly from boundary geography

[sgreenbury]

In terms of steps, it would require some edits to 3.2.2_assign_primary_zone_work.py:

1. Edit possible_zones_work before it is used here. The zones all match the boundary_geography resolution. If boundary_geography is OA, we should replace each possible_zones to it's parent MSOA and use that as possible_zones_work

2. After the assignment, each activity is mapped to an MSOA. If the boundary_geography is OA, we need to sample oan OA from the chosen MSOAs. This should be done before this step

Originally posted by @Hussein-Mahfouz in #52 (comment)

[Hussein-Mahfouz]

This is particularly useful if we are running on a big study area. The workzone assignment optimization problem might not be feasible at a high resolution (OA level) for a city like London. Running the problem at a lower resolution (MSOA level) would significantly reduce the number of variables.

One question to answer before implementing: Under what scenario would it be insufficient to run the entire pipeline at a lower resolution (e.g. MSOA level)?