Check RSpec's config.order when determining if a seed is used

[aprescott]

If a test suite has config.order = :random in its RSpec config, then the test ordering is random, but turbo_tests would not print a seed value. In order to get the correct behavior, a redundant --seed value needed to be given to turbo_tests.

The RSpec configuration of config.order = :random should now be correctly respected, so that running turbo_tests without --seed correctly prints the random seed used the same way that rspec does without giving --seed.

Note that this specifically does not try to handle the case where a custom ordering strategy has been given. Only the predefined strategy of :random is checked.

[aprescott]

This should address #60.

[aprescott]

As a meta note, I noticed some awkwardness in the repo while working on this change. Let me know if you'd like a separate issue for it:

The repo has a .rspec file. This file affects both the repo's own test suite and the exercised code under test. In other words, the 2 are coupled. For example, if you wanted to remove --color from .rspec to stop turbo_tests's own test output from including color, then tests would break because they rely on --color being present for bundle exec turbo_tests some_example_test_file_spec.rb. It might be better to clearly separate the repo's own test suite config from the config used by the spec files exercised by the test suite. (The latter shouldn't be so reliant on the shared global implicit state.)

[aprescott]

It occurred to me that the code here may not be sufficient. The actual config check logic is correct I belief, but the tests use an explicit --require to turbo_tests which doesn't actually prove an app's real config is in use. Moved to a draft PR for now.

[aprescott]

Can't get this working in a way that avoids problems with RSpec's global RSpec.configuration state, even in a happy path default case. (It ends up interfering with turbo_tests's own test config.) If there's a way to refer to the same RSpec config that RSpec itself would end up using, without modifying global state, then the diff here might be useful.