Add structured heading hierarchy to pkgdown references page

[Copilot]

The _pkgdown.yml was empty, causing pkgdown to render the references page as a flat alphabetical list with no organisation. This adds a proper heading structure and a data-raw script to keep it in sync as datasets are added over time.

Changes

_pkgdown.yml

Structured reference section with three top-level headings:

• Individual SSD datasets — five subheadings by source organisation:
  • Data from the Australian Institute of Marine Science (aims_*)
  • Data provided from anonymous sources (anon_*)
  • Data from the Australian and New Zealand guidelines (anzg_*)
  • Data from the Canadian Council of Ministers of the Environment (ccme_*)
  • Data provided by the Commonwealth Scientific and Industrial Research Organisation of Australia (csiro_*)
• Multi chemical SSD datasets — the five *_data combined datasets
• Derived SSD datasets (populated dynamically when anztox_data/wqbench_data are added)
• SSD Fit Results — ssd_fits
• Functions — ssd_data_sets, get_ssddata, gm_mean

data-raw/build_pkgdown.R

New script to regenerate _pkgdown.yml dynamically. Queries utils::data(package = "ssddata") at runtime so re-running it after new datasets are added (especially anzg_*) keeps the reference page accurate without manual edits.

# Run from package root
source("data-raw/build_pkgdown.R")
# → _pkgdown.yml written successfully.

The "Derived SSD datasets" section is omitted from the current file (no such datasets exist yet) but will appear automatically once anztox_data or wqbench_data land in the package.

[beckyfisher]

Thanks for putting this together — the structure and approach are very similar to what we landed on, which is a good sign. A few things in this PR needed correcting: the site URL was pointing to bcgov.github.io/ssdtools rather than open-aims.github.io/ssddata, getdata was missing from the Functions section, and the use of utils::data(package = "ssddata") for discovery requires the package to be installed rather than reading directly from man/. We've merged an alternative implementation in #22 that addresses these points and also adds the authors: block and development: mode: auto for the dev/release site split. Everything useful implemented in #20 is present in #22. This PR also makes some other minor tweaks to the documentation workflow. Closing this in favour of #22.