google-cloud-go: generate hook for checked-in postprocessing code (for BigQuery)

@shollyman writes: Our exemplar post-processing steps consist of Go code that is checked in and reviewed. It does not have unique dependencies beyond the go core library. It consumes the files generated by gapic-generator-go for a specific API service, and emits a generated Go source file. It is currently expected to be a different directory/package, but we could also make the argument it should live alongside the client definitions output by gapic-generator-go given the fact its another form of generated code.

The interesting questions are understanding how we define the post-processing flow for go.
* Should postprocessing code live in the go monorepo purely in source form, or should it be part of docker image?
* Should postprocessor tasks exist as a hermitic, compiled CLI, or should it satisfy some to-be-defined post-processor interface that's config driven?
* When a postprocessor task fails, it likely blocks generation. What does this mean in terms of a team "owning" a postprocessor task?