Update OCI model deployment clients to inherit from OpenAI

[qiuosier]

The OCI model deployments now supports OpenAI specs.

This PR removes the legacy implementation and implemented the new clients by inheriting from ChatOpenAI and OpenAI. Users will be able to use the parameters and methods supported by OpenAI.

Example:

from langchain_oci import ChatOCIModelDeployment
from langchain_core.messages import HumanMessage


llm = ChatOCIModelDeployment(
    model="odsc-llm",
    endpoint="https://modeldeployment.<region>.oci.customer-oci.com/<ocid>/predictWithResponseStream",
    max_tokens=256,
)

messages = [HumanMessage(content="Who's the first president of United States?")]
response = llm.stream(messages)

# Print the text of the response
for chunk in response:
    print(chunk.content, end="", flush=True)

The implementation requires langchain-openai library. However, this library is not required for users using the OCI generative AI clients.

[mrDzurb]

Maybe we can give some more user friendly message, that we rely on the langchain_openai and delegate inferencing to this library?