Add nvidia api

[Kipok]

import os
from nemo_skills.inference.server.model import get_model

os.environ['NVIDIA_API_KEY'] = "..."

model = get_model('openai', model="meta/llama3-70b-instruct", base_url="https://integrate.api.nvidia.com/v1")
prompts = [
    'Write a limerick about the wonders of GPU computing.',
]
model.generate(prompts)

or with a system message

import os
from nemo_skills.inference.server.model import get_model

os.environ['NVIDIA_API_KEY'] = "..."

model = get_model('openai', model="meta/llama3-70b-instruct", base_url="https://integrate.api.nvidia.com/v1")
prompts = [
    """<system_start>You're a coding machine, answer in Python only!<system_end>
       <user_start>Write a limerick about the wonders of GPU computing.<user_end>""",
]
model.generate(prompts)

[Kipok]

Merging without review - if any issues, let's fix them in follow-up PRs