[PROPOSAL] Remove waitCluster in Integration Tests

[nhtruong]

Every integration test starts of by hitting waitCluster method, which waits for the cluster to be up and ready before the test is executed. This is redundant because we have already implemented healthcheck along with autoheal to assure that the OS cluster is up and running before the integration workflow executes these tests.

Moreover, waitCluster is waiting for status of green, which is usually not applicable for local testing because:
Since we're testing the client functionality (mostly API endpoints), the OS cluster only has a single node (for simplicity and it's sufficient for development and testing). However, by default every index created requires 1 replica, which means it needs a 2nd node to deploy a backup copy to. This causes the index to have status of yellow, which leads to waitCluster to wait forever if there's any index already exists on the OS cluster. For most developers, it's normal to have a few existing indices on the local OS cluster for development. But we have to delete these indices or change their replicas_count to 0 if we want to run the integration tests locally.

Note that having a yellow status on single-node cluster is normal. The cluster is still working as intended. You just dont have a backup copy of the indices. I would say this is a bug from the server standpoint: Indices shouldn't have number of replicate set to 1 by default if there's only 1 node in the cluster.

In other words, we should remove waitCluster because:

• It's redundant
• It makes it harder to run integration tests locally

[nhtruong]

Workarounds to run integration tests locally:

• Remove all existing indices before running the tests: curl -X DELETE localhost:9200/_all
• Update settings of existing indices and set number_of_replicas to 0.

[dblock]

Workarounds to run integration tests locally:

• Remove all existing indices before running the tests: curl -X DELETE localhost:9200/_all
• Update settings of existing indices and set number_of_replicas to 0.

I suggest 1) documenting this in developer guide with something one can just copy-paste, 2) actually fixing integration tests to ignore existing data so that they can run on top of whatever is in an existing instance of OpenSearch (new issue?)

[nhtruong]

@dblock this issue has been resolved in #423. The workarounds were only meant for before #423 was merged.