when metadata is not embedded, query fails on metadata fields

[cmatKhan]

Including when using set_filter

> py$callingcards_hf$clear_filter("annotate_features")
> py$callingcards_hf$set_filter("annotated_features_meta" , data_usable="pass")
> callingcards_data = py$callingcards_hf$query("SELECT * FROM annotated_features", "annotated_features")
Query execution failed: Binder Error: Referenced column "data_usable" not found in FROM clause!
Candidate bindings: "target_symbol", "background_total_hops", "target_locus_tag", "callingcards_enrichment", "regulator_symbol"

LINE 1: SELECT * FROM metadata_annotated_features WHERE data_usable = 'pass'
                                                        ^
Final SQL: SELECT * FROM metadata_annotated_features WHERE data_usable = 'pass'
Error in py_call_impl(callable, call_args$unnamed, call_args$named) : 
  ValueError: Query execution failed: Binder Error: Referenced column "data_usable" not found in FROM clause!
Candidate bindings: "target_symbol", "background_total_hops", "target_locus_tag", "callingcards_enrichment", "regulator_symbol"

LINE 1: SELECT * FROM metadata_annotated_features WHERE data_usable = 'pass'
                                                        ^
Run `reticulate::py_last_error()` for details.

Potential solution: Create a 'view' of the joined metadata and data in duckdb, and query that rather than the parquet file itself.

Note: once that duckdb file exists, we could eliminate the parquet file in order to save space. The parquet file would only be re-downloaded if refresh is True and kept around long enough to update the view, then deleted again