duckdb-first version of package

[Aariq]

Given that analyses using all available fluxnet sites may be common, and that memory constraints prevent working with at least the daily and hourly data in-memory, it might be prudent to make the next version of this package work primarily with duckdb. There are now examples in https://github.com/EcosystemEcologyLab/fluxnet-annual-2026 of how to ingest all resolutions of data and a manifest into a duckdb database and how to check for and make updates.

I can imagine a workflow as follows:

1. flux_listall()
2. flux_download()
3. flux_extract()
4. flux_build_db(): constructs a duckdb database by ingesting available CSVs
5. flux_update_db(): compares flux_discover_files() with manifest stored in database and upserts data.
6. con <- flux_connect(): opens a connection to the duckdb database at a default location
7. annual <- tbl(con, "annual")
8. flux_qc(annual): returns a lazy tibble