Use a quasi-unique group name in this e2e test.

[3405691582]

I suspect there may be either some lack of cleanup or race that occurs when executing the tests, since the error mentions "An instance group already exists with name: cvd". I don't want to plumb too deeply into figuring out exactly why this happens; instead, just generate a quasi-unique group name based on the time and this way they should likely not collide any more.

[ser-io]

Let's fail if there's a cvd instance already created when the tests is about to run. The e2e test creating instances were implemented to be run in clean containers https://github.com/google/android-cuttlefish/blob/main/.github/actions/run-cw-sharded-e2e-test/action.yaml#L72 each of them.

[ser-io]

Explaining the group name collision error:

When flaky = True was added to snapshot_test_test bazel rule in #1934. Bazel started attempting the same test 3 times, as the test is known to be flaky, when the first attempt fails due the flakiness of cvd suspend with error: cvd suspend failed: execution of "/usr/bin/cvd" command with args ["/usr/bin/cvd" "--group_name=cvd" "--instance_name=1" "suspend"] failed: exit status 255, the next two attempt would fail in a deterministic way with error: An instance group already exists with name: cvd because the group was already taken in the first attempt and the test is meant to be run in a brand new environment every time.

[3405691582]

Different approach taken elsewhere.