Usage metering for embedding and reranking

[Yuqi-Du]

This PR will add ModelUsage as part the BatchedRerankingResponse.
The idea is to generalize ModelUsage that will be common shared by reranking and embedding model service.

public class ModelUsage 
  public final ProviderType providerType;
  public final String provider;
  public final String model;

  // requestBytes will be calculated according to the request http body to the provider.
  private int requestBytes = 0;
  
  // if content-length exists in provider http response, use that, otherwise, calculate bytes from http body in the response.
  private int responseBytes = 0;

  private int promptTokens = 0;
  private int totalTokens = 0;

Note that, gRPC prototype also changed to reflect the shared modelUsage structure.
So This PR will has corresponding EGW changes.

Checklist

• Changes manually tested
• Automated Tests added/updated
• Documentation added/updated
• CLA Signed: DataStax CLA

[amorton]

Picking this work up, and combining it with #1865

[amorton]

reviewed by yuqi