Add PPX support to the compiler

[jvanburen]

Adapt ppx_base to work in the compiler codebase. This will produce faster and smaller equality/comparison code than the hand-rolled ones, and it will make the codebase smaller and more readable.

[jvanburen]

this adds ppx_base as a dependency though, which we'll need to mention + update github actions...

[mshinwell]

I don't think we should do this, for at least the following reasons:

• there will never be acceptance of ppx upstream (correctly in my opinion) and so using it in flambda-backend is going to increase obstacles to upstreaming
• ppx is likely to make it harder to work on the compiler when it itself needs to be debugged, e.g. for the current arm64 warnings.ml problem. I believe this outweighs any concern about time spent manually maintaining these functions
• it will mean that the semantics of the compiler starts depending on the semantics of ppx
• the argument about the legibility and performance of the boilerplate functions isn't really a strong one in my opinion: in particular, it seems unlikely these are serious performance bottlenecks. Algorithms with poor time complexity or repeated IR traversals seem more likely suspects
• AI systems can now easily be used to write the boilerplate functions, which significantly reduces effort for new ones.

This seems likely to be related to #3914, but we should just fix the code generation if it is actually a problem.

[Gbury]

I agree with @mshinwell : I don't think the benefits of using/addping ppx in the compiler outweighs the (many) drawbacks it would inflict on us.