Add convenience functions for make_buffer to provide a shorthand for usage of the default mempools

[pciolkosz]

The following creates a buffer:

auto buffer = cuda::make_buffer(stream, cuda::device_default_memory_pool(dev), 42, 0);

As a shorthand we should provide:

auto buffer = cuda::make_device_buffer(stream, dev, 42, 0);

this way usage of default pools becomes more convenient and seems like the default option for the users

[miscco]

Question: how about cuda::make_device_buffer(stream, 42, 0); because the device is implicit in the stream