Saturation prior for PyMC-Marketing model with single geo still hierarchical

[TeemuSailynoja]

We need to fix the be3nchmarking.model_builder.build_pymc_saturation for single geo models. Now it sets a per geo hyper prior for each channel, which is likely to increase sampling time and increase posterior uncertainty.
What is even more important, that wouldn't really correspond to how an off-the-shelf model for that use case would be configured.