Support slice_head() and group_by() for occurrence downloads

[mjwestgate]

Our adoption of dplyr syntax opens up the possibility of more intuitive ways of calling the various APIs. It would also make the process more SQL-like, which would be more consistent with GBIF once we complete issue #249.

Two examples that spring to mind are:

1. supporting group_by() for occurrences. This would be a generalisation of atlas_species() to support any valid field as a facet; by so doing it would also resolve this issue: add rank argument to atlas_species() #195.

Example syntax:

galah_call() |>
  filter(year == 2024) |>
  group_by(class) |> # using species here would == atlas_species()
  collect()

2. supporting slice_head() for occurrences. This would require enabling the occurrences/search API for actual occurrence calls (currently only used for counts), and would require setting limits (as e.g. ALA only allows 500 records, from memory). This would help people view data before doing a large download.

Example syntax:

galah_call() |>
  filter(year == 2024) |>
  slice_head(n = 5) |>
  collect()

[nickdos]

@mjwestgate and I were discussing this issue in Slack and I suggested the Pipelines Spark interface might be a way of handling these requirements. This blog article provides some insights on how it works. GBIF experimental SQL downloads might be relevant too.

[mjwestgate]

Just flagging that supporting SQL downloads would not only solve this problem, but would also be consistent with our intention to move over to SQL for GBIF downloads in future (#249). Some discussion of options within the current API schema happened in biocache-service issue 829 in late 2023 but without deciding on an outcome.