fix: Ensured transactional safety when writing to redis

[sliverc]

Fixes #291

Added transactional safety when writing to redis to avoid corrupted data when multiple processes write to the database. This only works if the complete object is stored. As the redis pipeline itself stores the object quicker as it does it all in one go, I do not expect any performance issues through this change not to have dirty fields.

I would have loved to set up a test but was not successful to reproduce it in a unit test, as it not always occurs. But the tests still run like before, so confident that it does not break more than before.

@cunla Let me know what you think as you have a deeper inside into the project and whether this change makes sense from your point of view.

[sliverc]

To note with this change: it only fixes that no corrupt data is written into the redis database. In case if there is already invalid data in the redis database, the exception would stll be raised.

Either this data would need to be removed from redis directly or a PR implemented which ignores invalid models. I am happy to provide a PR in this direction but thought we first make sure this fix looks good and see how to move forward.