lint_package returns early when no package is found

[MichaelChirico]

On HEAD this gives me a funny error:

lintr::lint_package("/")
# Error in paths[rel_path] <- file.path(root, paths[rel_path]) : 
#   replacement has length zero

Should we issue a warning here?

[AshesITR]

Shouldn't this be an error condition?

[MichaelChirico]

An error seems harsh, though maybe the return should just be NULL?

Ran into this trying to create a test for #702 -- easiest test is to make sure a DESCRIPTION directory is skipped and then no package is found at all.

[AshesITR]

I would return NULL for find_package but lint_package() run non-interactively IMO should definitely error if there was no package, no?

If not, a warning ("No package was found at '/path'."?) with NULL instead of an empty lints list seems more correct.
Maybe even something like NA such that length(lint_package()) will be different from the lint-free state. WDYT?

[MichaelChirico]

I'm still happy with warn + return NULL. Don't really like NA -- the warning should be enough to distinguish from lint-free right?

Also it's NULL whereas lint-free is list().

[AshesITR]

LGTM. Please add a NEWS bullet.