Current: Each [Benchmark] constructs paths, reads from disk (data/Class.vb, data/Module.vb), and creates/configures VbaCompiler inside the timed method. This measures setup/IO more than compile performance.
Proposed: Shift path resolution and VbaCompiler construction/module/class registration into [GlobalSetup] or [IterationSetup]; leave the benchmark body to only call the compile method(s). Consider a separate benchmark if measuring setup is desired.