[mlir][gpu] Add metadata attributes for storing kernel metadata in GPU objects (#95292)

This patch adds the `#gpu.kernel_metadata` and `#gpu.kernel_table`
attributes. The `#gpu.kernel_metadata` attribute allows storing metadata
related to a compiled kernel, for example, the number of scalar
registers used by the kernel. The attribute only has 2 required
parameters, the name and function type. It also has 2 optional
parameters, the arguments attributes and generic dictionary for storing
all other metadata.

The `#gpu.kernel_table` stores a table of `#gpu.kernel_metadata`,
mapping the name of the kernel to the metadata.

Finally, the function `ROCDL::getAMDHSAKernelsELFMetadata` was added to
collect ELF metadata from a binary, and to test the class methods in
both attributes.

Example:
```mlir
gpu.binary @binary [#gpu.object<#rocdl.target<chip = "gfx900">, kernels = #gpu.kernel_table<[
    #gpu.kernel_metadata<"kernel0", (i32) -> (), metadata = {sgpr_count = 255}>,
    #gpu.kernel_metadata<"kernel1", (i32, f32) -> (), arg_attrs = [{llvm.read_only}, {}]>
  ]> , bin = "BLOB">]

```
The motivation behind these attributes is to provide useful information
for things like tunning.

---------

Co-authored-by: Mehdi Amini <[email protected]>

Author: fabianmcg

DialectGPU.cpp

@@ -48,17 +48,21 @@ PYBIND11_MODULE(_mlirDialectsGPU, m) {
       .def_classmethod(
           "get",
           [](py::object cls, MlirAttribute target, uint32_t format,
-             py::bytes object, std::optional<MlirAttribute> mlirObjectProps) {
+             py::bytes object, std::optional<MlirAttribute> mlirObjectProps,
+             std::optional<MlirAttribute> mlirKernelsAttr) {
             py::buffer_info info(py::buffer(object).request());
             MlirStringRef objectStrRef =
                 mlirStringRefCreate(static_cast<char *>(info.ptr), info.size);
-            return cls(mlirGPUObjectAttrGet(
+            return cls(mlirGPUObjectAttrGetWithKernels(
                 mlirAttributeGetContext(target), target, format, objectStrRef,
                 mlirObjectProps.has_value() ? *mlirObjectProps
+                                            : MlirAttribute{nullptr},
+                mlirKernelsAttr.has_value() ? *mlirKernelsAttr
                                             : MlirAttribute{nullptr}));
           },
           "cls"_a, "target"_a, "format"_a, "object"_a,
-          "properties"_a = py::none(), "Gets a gpu.object from parameters.")
+          "properties"_a = py::none(), "kernels"_a = py::none(),
+          "Gets a gpu.object from parameters.")
       .def_property_readonly(
           "target",
           [](MlirAttribute self) { return mlirGPUObjectAttrGetTarget(self); })
@@ -71,9 +75,16 @@ PYBIND11_MODULE(_mlirDialectsGPU, m) {
             MlirStringRef stringRef = mlirGPUObjectAttrGetObject(self);
             return py::bytes(stringRef.data, stringRef.length);
           })
-      .def_property_readonly("properties", [](MlirAttribute self) {
-        if (mlirGPUObjectAttrHasProperties(self))
-          return py::cast(mlirGPUObjectAttrGetProperties(self));
+      .def_property_readonly("properties",
+                             [](MlirAttribute self) {
+                               if (mlirGPUObjectAttrHasProperties(self))
+                                 return py::cast(
+                                     mlirGPUObjectAttrGetProperties(self));
+                               return py::none().cast<py::object>();
+                             })
+      .def_property_readonly("kernels", [](MlirAttribute self) {
+        if (mlirGPUObjectAttrHasKernels(self))
+          return py::cast(mlirGPUObjectAttrGetKernels(self));
         return py::none().cast<py::object>();
       });
 }