Resolves #453.

Much harder than I expected to implement, but I think my logic and implementation checks out. The algorithm is the following:

• Take the exclude set from CLI input and resolve it to actual nodes in the graph
• Determine all possible candidates for exclusion by performing DFS
  • So we add the direct and transitive dependencies of the exclude nodes we gathered in Step 1 into a candidates set, but we aren't sure yet if they should be excluded since a non-excluded node maybe using it. This is the reason why we have more steps below
• Create a reversed graph so that we know all the nodes that depend on a specific node
• Iterate over candidates, counting the number of dependents for each node using the reversed graph
• Now iterate over the explicitly-excluded nodes that we created in Step 1, where for each node:
  • Add it to the new excluded set
  • Iterate over their children. For each child, get the dependent count of that child and decrement it by 1
    • If the count equals 0, we haven't added it to the new set, and it isn't apart of the original resolved set created in Step 1, this child is a candidate for exclusion so let's add it (and in effect, also add their applicable children)
    • If this specific branch is never reached, it means that their is a non-excluded node that has an outgoing edge to child. We should not remove it
• Return the new excluded set, which will contain both the explicitly-excluded nodes and the dependency nodes associated to them