This separates the regeneration step from the build/install step. This is to simplify bootstrapping: trezorlib is needed in several build/test steps, and bound to be needed in more. It should be usable and installable straight from the checkout.
