assignment_linter() gains operator= argument to allow, e.g. '=' assignments

[MichaelChirico]

Supersedes #2698 and #2521. h/t again @J-Moravec for getting the ball rolling here.

Closes #2441.

This also un-couples the linted operator from the allow_trailing behavior; lints generated by the operator used are now independent of lints generated by line-trailing assignment operators:

## current 'main'
lint("a <-
1", assignment_linter(allow_trailing=FALSE))
# <text>:1:3: style: [assignment_linter] Assignment <- should not be trailing at the end of a line.
# a <-
#   ^~
lint("a =
1", assignment_linter(allow_trailing=FALSE))
# <text>:1:3: style: [assignment_linter] Use <-, not =, for assignment.
# a =
#   ^

## current branch (assignment-operator)
lint("a <-
1", assignment_linter(allow_trailing=FALSE))
# <text>:1:3: style: [assignment_linter] Assignment <- should not be trailing at the end of a line.
# a <-
#   ^~
lint("a =
1", assignment_linter(allow_trailing=FALSE))
# <text>:1:3: style: [assignment_linter] Use one of <-, <<- for assignment, not =.
# a =
#   ^
# <text>:1:3: style: [assignment_linter] Assignment = should not be trailing at the end of a line.
# a =
#   ^

Having <<- allowed by default is back-compatible, but does strike me as odd -- should we also plan to remove this from the defaults?

[MichaelChirico]

Thanks @IndrajeetPatil! I'd also like to get feedback from @AshesITR in this case since he commented a lot on the earlier PRs & this will be released soon afterwards.

[IndrajeetPatil]

Having <<- allowed by default is back-compatible, but does strike me as odd -- should we also plan to remove this from the defaults?

Yeah, I think it'll be good to remove it from the defaults.

[AshesITR]

Good stuff, just a few kinks that could be ironed out.

[AshesITR]

Positional changes could affect downstream dependencies if the linter is configured using positional arguments. We should at least check for is.logical(operator) to stop with a helpful error message.

[AshesITR]

This path will produce a confusing message for operator = c("<-", "<<-") when encountering "->>".

[AshesITR]

This will be inconsistent advice if operator = "=".

[AshesITR]

For the test structure, maybe check for deprecation warnings once (maybe with smell tests that the automatic inference of operator works correctly) and proceed testing only the new spec with all (old) tests?

That should make full removal easier down the line.