Dynamic run-time parquet loading

[jpambrun]

Description

I want to be able to load parquet dynamically. The source parquet will change often and the rebuild workflow is not really workable. I also would like to be able to dynamically change parquet.

What problem would this solve?

Allow realtime-ish data update or dynamic parquet source.

How should it work?

I was trying to hack it like this

<script>
  import { setParquetURLs,updateSearchPath } from '@evidence-dev/universal-sql/client-duckdb';
  setParquetURLs( { schema: ["http://localhost:5000/outputs_wo.parquet"] } );
</script>

``sql models
  select
      model
  from schema.outputs_wo
  group by model
``

It worked the first time so I was happy, but there is race condition and in reality is works extremely rarely. I would need to be able to await setParquetURLs (with the queries delayed), but the script tag appeared to be inserted into a non-async function.

I think it would be really nice to support this. We could then use query params to select an appropriate parquet.