HDDS-16365. Pipeline.getReplicaIndexes() unnecessarily create a new map.

[szetszwo]

What changes were proposed in this pull request?

In Pipeline, the getReplicaIndexes() method, which copies the replicaIndexes map, is not really needed:

1. In many cases, the copied map is used for calling get(dn) -- just directly call the existing getReplicaIndex(dn) method.
2. One special case is in the ScmClient in OM for checking if a pipeline containing all replica indexes -- create a new containsAllReplicaIndexes method.
3. The remaining cases are used in tests for iterating the nodes -- rename it to getReplicaIndexesForTesting() and use unmodifiableMap instead of copying.

For #1 above, getReplicaIndexes() is mainly used for building DatanodeBlockID protos. We will also refactor the code:

• Add a new replicaIdx parameter to BlockID.getDatanodeBlockIDProtobufBuilder().
• For the code using DatanodeBlockID.Builder, change them to use getDatanodeBlockIDProtobufBuilder(..) instead.

What is the link to the Apache JIRA

HDDS-16365

How was this patch tested?

By updating existing tests.