Motivation

With recent refactorings to namespace packaging (#11190 ) and usage of setuptools_scm (#11355) the kubernetes dev run script broke.
It still mounted pro source code into site-packages/localstack-ext and not inside the localstack namespace package, and the entrypoint mounts did not take into account the new versioning scheme.

A main culprit was the implicit behavior of taking parts of the path from the host and using them inside the container, as this prevented mounting the pro source code in the localstack/pro directory.
Changing this required some refactoring.

Changes

• The mount points are now generated in a separate method, and into a defined data structure. This provides more flexibility, as previous behavior was very implicit (taking the last part of the path on the host for mounting into the container for example). The config generation methods use the data and convert it into the respective format, without additional logic.
• Add --mount-entrypoints flag, to manually opt in to mounting the entrypoints. I did not yet implement a version detection to mount the entrypoints into the right .egg-info directory, so this has to be manually adjusted currently for the versions inside the docker image, if used.
• Add CONTAINER_RUNTIME=kubernetes and lambda.executor=kubernetes config to the value overrides if pro is enabled
• Fix some errors when mounting code into the community package (as the package installation of localstack-core differs in community from pro)

Testing

Run python -m localstack.dev.kubernetes --pro --write --output-dir=<config-dir> and follow the instructions.