Personally I am using an architecture similar to https://sive.rs/pg in my personal projects and I am very happy with it. It is important that you put data and procedures/views in different schemas so that you can use migrations for your data but automatically delete and recreate the procedures/views during deployment. Also use pgtap or something similar for testing.
Example code: https://github.com/sivers/store
This talk is more Oracle specific but with some minor changes you can apply the same ideas to postgres: https://www.youtube.com/watch?v=GZUgey3hwyI
Personally I am using an architecture similar to https://sive.rs/pg in my personal projects and I am very happy with it. It is important that you put data and procedures/views in different schemas so that you can use migrations for your data but automatically delete and recreate the procedures/views during deployment. Also use pgtap or something similar for testing.