Support for flatgeobuf format

[jaredlander]

Prework

• Read and agree to the code of conduct and contributing guidelines.
• If there is already a relevant issue, whether open or closed, comment on the existing thread instead of posting a new issue.
• New features take time and effort to create, and they take even more effort to maintain. So if the purpose of the feature is to resolve a struggle you are encountering personally, please consider first posting a "trouble" or "other" issue so we can discuss your use case and search for existing solutions first.
• Format your code according to the tidyverse style guide.

Proposal

When a target results in a large object the standard 'rds' format can be slow. For data.frames and alike alternatives like 'fst' and 'parquet' are pretty great. However, sf objects don't work well with that format (it's surprising that 'parquet' doesn't play nicely). Even 'qs' gives me trouble.

So I'd love to see a 'flatgeobuf' option. The {sf} package supports writing to flatgeobuf by calling sf::sf_write(x, 'file_name.fgb') if gdal > 3.1 is installed (I can't get that installed on my machine for some reason so I wasn't able to test this myself).

My thinking would be that setting format='flatgeobuf' would write the sf object to this type of file quickly.

list(
    tar_target(
        one_point,
        tibble::tibble(ID=1:2, X=c(54.24, 52.34), Y=c(34.17, 33.31)) %>% 
            sf::st_as_sf(coords=c('X', 'Y')),
        packages=c('magrittr'),
        format='flatgeobuf'
    )
)

I tried doing this myself for a PR my copying the parquet format, but the wrong version of gdal is stopping me (and I know I need to fix that before I could even use this feature).

[wlandau]

I would still prefer other solutions than a built-in format for this. I try to keep formats general: data frames, popular machine learning objects, and R objects in general. sf is popular, but it is still domain-specific enough to give me pause.

I also feel like arrow is really close to being able to just save sf objects as feather or parquet files. What have the maintainers of those packages said about this? If either sf or arrow could bridge the gap, then it would benefit more R users than just targets users.

[wlandau]

One extra format does not seem like much at first, but I have started to add "aws_*" versions of all formats, and I am considering Google Cloud and Azure. Those are the major players in the cloud right now, but even more could emerge. Local formats X cloud providers is already enough combinatorics to be dangerous.