[torch-mlir][sparse] add sparsification to linalg reference backend (#2887)

This adds a few passes that will ensure linalg with sparse tensors are
properly lowered to loops and can run using the ExecutionEngine for
testing (a few details on parameter passing from PyTorch still TBD)

Test results:

$ ./tools/e2e_test.sh --config linalg

Summary:
    Passed: 1144
    Expectedly Failed: 8

$ python -m e2e_testing.main --config=torchdynamo -v

Summary:
    Passed: 960
    Expectedly Failed: 163

Filed issue:
https://github.com/pytorch/pytorch/issues/119407

projects/pt1/e2e_testing/xfail_sets.py

@@ -35,6 +35,11 @@ TORCHDYNAMO_XFAIL_SET = {
     "ConvolutionBackwardModule2DPadded_basic",
     "ConvolutionBackwardModule2D_basic",
 
+    # Size result mismatch (exposed by downstream canonicalizer
+    # on incompatabile casts).
+    # https://github.com/pytorch/pytorch/issues/119407
+    "ConvolutionBackwardModule2DStrided_basic",
+
     # RuntimeError: Index tensor must have the same number of dimensions as self tensor
     # RuntimeError: Failed running call_function aten.nll_loss_backward(...
     # https://github.com/pytorch/pytorch/issues/89630

projects/pt1/python/torch_mlir_e2e_test/linalg_on_tensors_backends/refbackend.py

@@ -131,8 +131,13 @@ LOWERING_PIPELINE = "builtin.module(" + ",".join([
     # This is likely because if things are naturally fusable we usually already
     # emit things in that form from the high level (e.g. single linalg-generic).
     # Other backends are likely to benefit more.
+    "func.func(linalg-generalize-named-ops)",
     "func.func(linalg-fuse-elementwise-ops)",
     "convert-shape-to-std",
+    # MLIR Sparsifier mini-pipeline. Note that this is the bare minimum
+    # to ensure operations on sparse tensors are lowered to loops.
+    "sparsification-and-bufferization",
+    "sparse-storage-specifier-to-llvm",
     # Bufferize.
     "func.func(scf-bufferize)",
     "func.func(tm-tensor-bufferize)",
@@ -199,7 +204,9 @@ class RefBackendLinalgOnTensorsBackend(LinalgOnTensorsBackend):
 
         run_pipeline_with_repro_report(
             imported_module, LOWERING_PIPELINE,
-            "Lowering Linalg-on-Tensors IR to LLVM with RefBackend")
+            "Lowering Linalg-on-Tensors IR to LLVM with RefBackend",
+            enable_ir_printing=False,
+        )
         return imported_module
 
     def load(self, module) -> RefBackendInvoker: