Ship the merged CPU kernels as their own library

[shoumikhin]

The CPU operator kernels are compiled into whichever component links
them. The Python extension therefore carries a full private copy, which makes it large, and a
C++ application cannot get the same operators without building them from source.

before:
  _portable_lib.so     contains its own copy of every CPU kernel
  your C++ app         cannot get them without building from source

after:
  executorch/lib/libexecutorch_optimized_native_cpu_ops_lib.so.1
    _portable_lib.so   -> links it
    your C++ app       -> links it

There is a correctness reason too, not just size. Operators register themselves into a table
when their library loads. Two copies in one process means two independent registrations of the
same operator set, and which one a model resolves to is then a matter of load order.

How you use it

find_package(executorch REQUIRED)
target_link_libraries(my_app PRIVATE executorch::runtime executorch::kernels)

The target name is the contract, not the file name. A model running the usual CPU operators
needs nothing else.

Tested

Built the wheel on Linux x86_64 and aarch64, installed it into a clean environment, and
confirmed by symbol inspection that exactly one shipped library registers the CPU operator set.
Linked a C++ application against the component, ran a model through it, and ran the existing
Python tests so current users are unaffected.

[shoumikhin]

Stack from ghstack (oldest at bottom):

• Check the shipped device code covers the GPUs a wheel claims #21567
• Publish CUDA wheels so a GPU user does not build from source #21569
• Ship the CUDA delegate as its own library #21531
• Ship the XNNPACK delegate as its own library #21526
• -> Ship the merged CPU kernels as their own library #21524
• Ship the thread pool as its own library so a process has one pool #21522
• Ship a linkable shared runtime and make the Python bindings use it #21514

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21524

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.