[MLIR][KERNEL_GEN] Add a library to lower kernels with the host side.

* Unified TF->Cubin and TF->Kernel_with_host side lowering in `kernel_creator.h|cc`
* Added a pass that attaches GPU binary blob to GPUModuleOp
* Refactored most of the code.
* Added tf_to_kernel binary that emits obj file

PiperOrigin-RevId: 330494488

include/mlir-hlo/Dialect/mhlo/transforms/passes.h

@@ -53,7 +53,7 @@ std::unique_ptr<OperationPass<ModuleOp>> createLegalizeToLhloPass(
 std::unique_ptr<OperationPass<FuncOp>> createLegalizeHloToLinalgPass();
 
 // Transforms unranked HLO operations to ranked ones where possible.
-std::unique_ptr<OperationPass<FuncOp>> createTransformUnrankedHloPass();
+std::unique_ptr<FunctionPass> createTransformUnrankedHloPass();
 
 // Sinks constants implicitly captured in control flow regions. This is
 // necessary to export to XLA.

lib/Dialect/mhlo/transforms/transform_unranked_hlo.cc

@@ -198,7 +198,7 @@ void PopulateTransformUnrankedHloPatterns(MLIRContext *context,
   // clang-format on
 }
 
-std::unique_ptr<::mlir::Pass> createTransformUnrankedHloPass() {
+std::unique_ptr<FunctionPass> createTransformUnrankedHloPass() {
   return std::make_unique<TransformUnrankedHloPass>();
 }