inclusive_scan_by_keys calls the 'sum' operator needlessly

inclusive_scan_by_keys is called with a one-element keys vector. It should just copy the first input element to the output vector but in addition it invokes the 'sum' operator 4096 times!

Reproduction code (bug.cu)
```
#include
#include

#include

class Agg {
public:
__device__ int operator()(int x, int y) {
printf("%d %d\n", x, y);
return x + y;
}
};

int main(int argc, char** argv) {
auto partitions = thrust::device_vector(1, 0);
auto in = thrust::device_vector(1, 0);
auto out = thrust::device_vector(1, 0);

thrust::inclusive_scan_by_key(
partitions.begin(),
partitions.end(),
in.begin(),
out.begin(),
thrust::equal_to(),
Agg()
);
return 0;
}

```

Build
```
nvcc --std=c++14 -o bug bug.cu
```

Environment
```
CentOS 7 (Kernel 3.10.0-1160.el7.x86_64)

NVIDIA-SMI 470.82.01
Driver Version: 470.82.01
CUDA Version: 11.4

GPU: NVIDIA GeForce RTX 2060
```