server: add inference path option to make API compatible with openai client

[eschmidbauer]

This change makes it possible to use the openai client SDK

./server --host 0.0.0.0 --port 9010 -nt \
     -m models/ggml-large-v3-q5_0.bin --request-path /audio/transcriptions --inference-path ""

from openai import OpenAI

client = OpenAI(api_key="1", base_url="http://localhost:9010")

with open("test.wav", "rb") as file:
    payload: bytes = file.read()
    transcription = client.audio.transcriptions.create(
        model="models/ggml-large-v3-q5_0.bin",
        file=file
    )
    print(transcription.text)