Proposal: move op enum generation into build process

[Dewb]

The build process currently consumes op_enums.h from source control; this is a generated file that must be manually regenerated if src/ops/op.c changes. Proposal: remove op_enums.h from source control, add it as a target to src/module/Makefile and src/simulator/Makefile, and run the Python script to generate it during the build.

Pros:

• The developer documentation could be simplified
• Codegen output would always be current, no chance of forgetting to run the script when necessary
• Would be easier to change codegen strategies in the future without making devs relearn new manual steps

Cons:

• The generated IDs in op_enums.h would no longer be visible/searchable in GitHub
• Python (and/or any other future codegen dependencies) would always be a requirement for building the firmware, instead of just being a requirement for adding new ops
• Build process would get a little bit longer

[Dewb]

I think the pros outweigh the cons; folks already benefit from being steered to local-search investigations because GitHub does not search the libavr32 submodule; if devs are building with the Docker image, the dependencies are taken care of, and the build process already depends on Ragel; the build time increase should be quite small.

[tehn]

Agreed pros outweigh cons.

[Dewb]

This is 95% done, I just have to test on hardware.