Add config_file = argument to lint() & co.

[AshesITR]

Would be more convenient than withr::with_options(c("lintr.linter_file" = my_new_config), lintr::lint_package()) when prepairing a config switch.

[MichaelChirico]

WDYT about re-using parse_settings for this?

lint_package(parse_settings = my_new_config)

[AshesITR]

Good idea WDYT about renaming to config = NULL and accepting:

1. NULL or TRUE for auto-discovery
2. NA or FALSE for parse_settings = FALSE
3. A string path to a .lintr or .lintr.R
4. A list containing a constructed config

Of course we would need a deprecation cycle and some way to detect DCF vs. R.

[MichaelChirico]

I'm not personally a fan of requiring people to distinguish NULL from NA, it always feels a bit arbitrary. E.g. I'd assume NULL --> "nothing" --> "no config", vs. NA --> "missing" --> ?not sure

Besides that the idea sounds good initially. I would use config = "auto" the default for auto-discovery. We could also let config = "path/to/dir" where the directory is searched for .lintr / .lintr.R as today.

[MichaelChirico]

Related: #1515

[AshesITR]

The issue with "auto" is a surprising edge case four the user if the desired config is located in a directory with name "auto".

Using NULL as not specified / default seems less surprising to me.
I agree we shouldn't act on NA though.