Removed unused DuckDB code path, dead helpers, and 16 MB unused data

[eboyer221]

Description

PR removes unused data files and dead code that is no longer necessary for amRviz functionality.

The dashboard no longer uses inst/extdata/amr_filtered_tbls.db (16MB), it was superseded by the parquet-based loading path that landed in previous PRs. This PR removes that file as well as functions that are no longer needed in order to free up space and clean the codebase.

Removed

Functions (all in R/utils.R, all with zero references outside their own definition — verified by grep across R/, inst/, tests/):

• getAllFilteredData — old DuckDB writer; populated the .db from per-species DuckDBs.
• loadAMRMetadata — old DuckDB reader; the dashboard now uses loadMLResults / loadTopFeat / get_metadata_path reading parquet files instead.
• getTopFeatures — pre-parquet loader for top-features TSVs that wrote into a shinyapp/data/ path that doesn't exist in the package.
• combinePerformanceMetrics — same pre-parquet pipeline; wrote to the same non-existent shinyapp/data/ paths.
• getAmrDrugs
• bug_choices - hardcoded ESKAPE lookup. The dashboard now derives bug choices dynamically from loaded data (this was finalized in PR Fix install + dashboard rendering with bundled demo data #18).

Data: inst/extdata/amr_filtered_tbls.db (16MB). Only the deleted functions touched it.

Imports: DBI and duckdb removed from DESCRIPTION and the matching @importFrom roxygen tags in R/utils.R. The only callers of DBI::* and duckdb::* were inside the deleted functions, so the package no longer depends on either.

README: removed the amr_filtered_tbls.db bullet under "Data requirements" - that line was already inaccurate since the dashboard hasn't read from the DB in some time.

• inst/extdata/ now contains only the per-species parquet files and the small TSV/CSV lookups (≪1MB total).

Test plan

1. devtools::load_all() then launchAMRDashboard() - dashboard builds and every tab loads with no missing-function errors.
2. devtools::check() - no errors and no test failures introduced.
3. Verification grep: `grep -rn "DBI|duckdb|amr_filtered_tbls|bug_choices|getTopFeatur

[AbhirupaGhosh]

Just a suggestion, we can switch to the base R pipe (|>) throughout, and only retain %>% where it's explicitly required.

[eboyer221]

That's a good point @AbhirupaGhosh. I could get rid of magrittr from our Imports if that switches. I assume that would be better for Bioconductor, since they don't want unnecessary dependencies?

@amcim @jananiravi - Should I do a separate follow-up PR that migrates the package to |> and drops magrittr from Imports?

[amcim]

Yes it would definitely be worth making that conversion to |> and it will be a quick PR to review