[MLIR][Shape] Limit `shape.rank` lowering to its extent tensor variant

[frgossen]

Summary

When lowering to the standard dialect, we currently support only the extent
tensor variant of the shape.rank operation. This change lets the conversion
pattern fail in a well-defined manner.

Depends On D84430

[rriddle]

Could you please add a test case?

[frgossen]

See https://reviews.llvm.org/D84852

mlir/lib/Conversion/ShapeToStandard/ShapeToStandard.cpp

matchAndRewrite(shape::RankOp op, ArrayRef<Value> operands,
                   ConversionPatternRewriter &rewriter) const override;
 };
 } // namespace
 
 LogicalResult
 RankOpConverter::matchAndRewrite(shape::RankOp op, ArrayRef<Value> operands,
                                  ConversionPatternRewriter &rewriter) const {
   shape::RankOp::Adaptor transformed(operands);
-  rewriter.replaceOpWithNewOp<DimOp>(op, transformed.shape(), 0);
+  Value shape = transformed.shape();
+
+  // For now, only extent tensor lowering is supported.
+  if (shape.getType().isa<ShapeType>())
+    return failure();
+
+  rewriter.replaceOpWithNewOp<DimOp>(op, shape, 0);
   return success();
 }
 
 namespace {
 /// Type conversions.
 class ShapeTypeConverter : public TypeConverter {
 public:
   using TypeConverter::convertType;