Use non-mangled type names in nvbench output (#16649)

Uses the `NVBENCH_DECLARE_TYPE_STRINGS` feature to produce readable type names in the nvbench output.
Example previous output for `cudf::timestamp_ms` would appear like this:
```
| cuda::std::__4::chrono::time_point<cuda::std::__4::chrono::system_clock, cuda::std::__4::chrono::duration<long, cuda::std::__4::ratio<1l, 1000l> > > | 100000 |  23840x | 25.138 us | 21.98% | 20.979 us |  9.54% | 4.767G |  38.134 GB/s |  4.38% |
```
Adding the nvbench name feature changes this to:
```
| cudf::timestamp_ms | 100000 |  24752x | 24.387 us | 21.58% | 20.208 us | 3.79% | 4.948G |  39.588 GB/s |  4.55% |
```

Authors:
  - David Wendt (https://github.com/davidwendt)

Approvers:
  - Bradley Dice (https://github.com/bdice)
  - Mark Harris (https://github.com/harrism)

URL: #16649

cpp/benchmarks/reduction/minmax.cpp

@@ -47,6 +47,8 @@ static void reduction_minmax(nvbench::state& state, nvbench::type_list<DataType>
   set_throughputs(state);
 }
 
+NVBENCH_DECLARE_TYPE_STRINGS(cudf::timestamp_ms, "cudf::timestamp_ms", "cudf::timestamp_ms");
+
 using Types = nvbench::type_list<bool, int8_t, int32_t, float, cudf::timestamp_ms>;
 
 NVBENCH_BENCH_TYPES(reduction_minmax, NVBENCH_TYPE_AXES(Types))

cpp/benchmarks/reduction/reduce.cpp

@@ -81,6 +81,8 @@ static void reduction(nvbench::state& state, nvbench::type_list<DataType, nvbenc
   set_throughputs(state);
 }
 
+NVBENCH_DECLARE_TYPE_STRINGS(cudf::timestamp_ms, "cudf::timestamp_ms", "cudf::timestamp_ms");
+
 using Types    = nvbench::type_list<int32_t, int64_t, double, cudf::timestamp_ms>;
 using AggKinds = nvbench::enum_type_list<cudf::reduce_aggregation::MIN,
                                          cudf::reduce_aggregation::SUM,