Fix is_device_write_preferred in void_sink and user_sink_wrapper

[vuule]

Description

Addresses a few issues in data_sink classes to avoid D2H copies in writers when using a void_sink.
Provide an is_device_write_preferred implementation to always prefer device writes.
Implement is_device_write_preferred in user_sink_wrapper that forwards the call to the wrapped object.
Use the cudf::io::void_sink in benchmarks instead of the local version, which is not fully implemented.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[vuule]

This should speed up some of the writer benchmarks significantly (~2X), but does not improve throughput in any real use case.
The only possible impact might come from fixed is_device_write_preferred in user_sink_wrapper, but we don't cover user implemented sinks in libcudf benchmarks.

[ttnghia]

/merge