[Release] Fix binary name for downstream compatibility

[sjain-stanford]

As of Sep 14, the torch-mlir binary wheels got renamed to torch-mlir-core from torch-mlir:

This was an unintended side-effect of the recent change of TORCH_MLIR_ENABLE_ONLY_MLIR_PYTHON_BINDINGS=True (#3711) which skips setting NAME = "torch-mlir" in setup.py.

To avoid having multiple downstreams fix their pip deps, this change allows using the same torch-mlir name for binaries, and reserves a separate torch-mlir-ext name for the (less popular) binaries with extensions enabled.

[saienduri]

Fine by me, as torch-mlir-core seems to be the main thing that we support and test right now so makes sense to change to torch-mlir. Can you also change all the other occurrences of these package names in the repo to torch-mlir and torch-mlir-ext so we have uniform naming convention? Let's also wait for Stella's opinion (she's probably the one that named them initially lol)

[sjain-stanford]

Can you also change all the other occurrences of these package names in the repo to torch-mlir and torch-mlir-ext so we have uniform naming convention?

@saienduri I have done this now, but just for the record, until now we were really calling the release builds with PT extensions (TM_PACKAGES=torch-mlir) here - which ran build_torch_mlir.

After this change, we will actually be building the "core" (non-PT-extension) version. It is unlikely this breaks anything downstream for y'all but if it does, we could simply switch this line in the GHA workflow to go back:

- TM_PYTHON_VERSIONS=cp310-cp310 TM_PACKAGES=torch-mlir ./build_tools/python_deploy/build_linux_packages.sh
+ TM_PYTHON_VERSIONS=cp310-cp310 TM_PACKAGES=torch-mlir-ext ./build_tools/python_deploy/build_linux_packages.sh

I'm going to go ahead and land this. 🤞🏻