gcs provider fixes

[jbardin]

• The backend was creating the bucket named in the configuration if it didn't exist. We don't allow other backends to do this, because these are not managed resources that terraform can control.
• Have the test fixtures create their own buckets, so parallel tests don't collide, and leftover buckets don't cause future failures. This also properly removes test buckets regardless of contents.
• Add a test that fails the condition that This change will fix the dobule lock issue when we use data.terrafo… #16852 was fixing. This is simply looking up a locked remote state, since a backend may need to take a lock to create a non-existent state. Reading an existing locked remote state should succeed.
• Match the operation of the google provider, by accepting either a file path or contents for both credentials and GOOGLE_CREDENTIALS

[jbardin]

$ TF_ACC=1 go test -v 2>/dev/null
=== RUN   TestStateFile
=== RUN   TestRemoteClient
=== RUN   TestRemoteLocks
=== RUN   TestBackend
=== RUN   TestBackendWithPrefix
--- PASS: TestStateFile (0.00s)
--- PASS: TestRemoteClient (6.73s)
--- PASS: TestRemoteLocks (8.95s)
--- PASS: TestBackend (18.68s)
	backend_test.go:114: TestBackend: testing state locking for *gcs.gcsBackend
--- PASS: TestBackendWithPrefix (21.86s)
	backend_test.go:127: TestBackend: testing state locking for *gcs.gcsBackend
PASS

[jbardin]

This also fixes #16803

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.