Add parquet format as an output for embeddings?

The [Embedding Explorer](https://github.com/Imageomics/emb-explorer) (and [image search](https://github.com/Imageomics/bioclip-image-search-lite)?) uses a Parquet database for input of vision embeddings. Maybe we should offer the option here to generate Parquet right here to reduce friction from getting from pybioclip output to Embedding Explorer input.

@netzissou thoughts? Is there a spec for required columns?

Start with the pybioclip CLI output path and inspect the Parquet input requirements of Embedding Explorer and image search. Confirm the required columns and output option with the issue participants; done means pybioclip can produce Parquet accepted by the referenced tools.

Written by the indexing model from the issue text.