[Fix #12975] Fix an error for gem_config_path when bundler is loaded and no Gemfile exists

[Earlopain]

Fixes #12975

rubocop/lib/rubocop/lockfile.rb

Lines 3 to 10 in 10d115f

	begin
	# We might not be running with `bundle exec`, so we need to pull in Bundler ourselves,
	# in order to use `Bundler::LockfileParser`.
	require 'bundler'
	rescue LoadError
	nil
	end

Bundler may be loaded because the LockFile parsers needs it.

With the improvement to the isolated environment, 4 tests fail. Previously they operated on the gemfile from the project folder.

The two autogen tests relied on modified env, which now is not visible anymore.

---

This uses Bundler.with_unbundled_env because bundler adds some variables that would otherwise make this hard to test. Bundler.reset! is needed because a bunch of paths are memoized which would cause bundler to always return the Gemfile path of the rubocop folder instead of the tempfile/pwd. I added a test for that.

---

Before submitting the PR make sure the following are checked:

• The PR relates to only one subject with a clear title and description in grammatically correct, complete sentences.
• Wrote good commit messages.
• Commit message starts with [Fix #issue-number] (if the related issue exists).
• Feature branch is up-to-date with master (if not - rebase it).
• Squashed related commits together.
• Added tests.
• Ran bundle exec rake default. It executes all tests and runs RuboCop on its own code.
• Added an entry (file) to the changelog folder named {change_type}_{change_description}.md if the new code introduces user-observable changes. See changelog entry format for details.

[Earlopain]

:/ this might isolate a bit too much actually. Not quite sure why it passes locally.

I think it would be nice to improve isolation like this here since I would expect it to work like this already. Any ideas on that? Otherwise I will just make it isolate like that once for a test that needs this (and hope that passes on CI)

[Earlopain]

Sigh, bundler really wants to keep its values around somewhere. I openend #12978 for the general problem but even just this single test ended up more complicated than I really wanted it to. It does test the correct thing at least.

[bbatsov]

Thanks for tackling this!