This is a performance optimization when using the --stacks/--targets flag, and extracted from #632.

Think of the following example

stacks:
  - name: a
    locked: true
  - name: b
    requires: [a]
    locked: true
  - name: c
    requires: [b]

This forms a dependency graph like so:

c -> b -> a

Given I run stacker build --targets c, stacker will walk all 3 nodes in the graph, however, given that b is locked, it's unnecessary for stacker to execute a; since b is locked it doesn't need any information from a, and thus, doesn't actually depend on a.

With this change, because b is locked, it has no dependencies, so a will not be included in the filtered graph, which becomes:

c -> b