Usually the problem is not the bloat, but the management of the feature and the expectations of users post-introduction. I’m not saying this would be the case, but existing a more complete solution to do it (and with a developer fully dedicated to it), I think it’s better to use that and keep just a minimal feature set in darktable.
I would argue that having two pieces of software that do two very different things in an efficient way is better than two half-implemented features in a single software. But each one has it’s opinions on that. 