Safer check for NAs in tiledb_config()

[mojaveazure]

Replace if(is.na()) as R 4.2 errors when conditions > 1L Instead, removes NAs from config
Then, checks the length of config; if there is a length, adds it to the TileDB config

fixes #518

[eddelbuettel]

This looks good. It may be a function that predates me, and the use if NA here is a little non-standard (maybe check for missing() or empty string is cleaner) but changing it now and changing the interface is not great either.

[eddelbuettel]

Looks good especially with the added tests.

[mojaveazure]

missing has it issues, so I tend to prefer having a default over using missing; Seurat uses NULL since rlang has %||% to check for it, but I think NA works well enough here. Plus, in the help page, the default is NA_character_ which indicates that config should be a character

[eddelbuettel]

Let's table whether rlang is authorative for another day but I guess I meant to approve and comment and just commented. Correcting that now.

[eddelbuettel]

The other reason that issue went undetected is that most uses of config use this pattern for one or more settings

cfg <- tiledb_config()
cfg["vfs.s3.region"] <- "us-west-1"
ctx <- tiledb_ctx(cfg)

and the more compact form of the named charecter vector which, as you correctly identified could be longer than length one, is not use much (but even more convenient).

[eddelbuettel]

Oh, one more thing: can you add a line to NEWS.md with the PR number? I found adding [ci_skip] to the commit makes it a little lighter on non-code commits.

And then as it is approved feel free to squash-merge.

[eddelbuettel]

Thanks for updating the file. Per precedent in this file (as well as what Core does) I tend to only list the PRs, not issues. We can address that in the next edition, no rush.