Implement native bulk importers where possible

[cboettig]

Ideally, arkdb should detect when the input file is compatible with a known bulk importer method and then use that to import rather than importing in chunks.

For instance, MonetDB and duckdb have quite reasonable bulk importers. Others can at least do csv. In most cases this should be much faster than chunking, though tends to be less flexible and will need a mechanism to override

[cboettig]

#24, #27 implement an example for MonetDBLite.