Remove prims.embedding and prims.embedding_backward

[IvanYashchuk]

We don't need prims.embedding and prims.embedding_backward because we can simply use thunder.torch.embedding and thunder.torch.embedding_backward as our transformation targets. When we added these primitives we didn't have a good idea how multi-level symbols can be targeted.

With this PR thunder.torch.embedding is transformed directly to torch.nn.functional.embedding for execution. While before this PR thunder.torch.embedding is decomposed into prims.embedding which is then transformed into torch.nn.functional.embedding (and similarly for embedding_backward).

Testing:

pytest thunder/tests/test_grad.py -k "embedding" -vvv
pytest thunder/tests/test_ops.py -k "embedding" -vvv

[mruberry]

This is interesting.

Isn't this regressing support for calling torch.embedding with the max_norm argument, though?

[IvanYashchuk]

The max_norm argument was not supported. I don't know the reason for this. I think it should work and I'll update OpInfo samples to test it.

[IvanYashchuk]

max_norm argument when active modifies the weight argument inplace. That's probably the reason this argument raised NotImplementedError. I'll keep the behavior as is in this PR.
Created an issue for tracking completion #1699.

[riccardofelluga]

A wait then we might want to change this too

lightning-thunder/thunder/executors/nvfuserex_impl.py

Lines 2636 to 2637 in 52ee541

	register_supported(PrimIDs.EMBEDDING, embedding, _embedding_check)
	register_supported(ltorch.embedding, embedding, _embedding_check)

cc. @Priya2698 for #1674