feat: group deps by provider

[cdoern]

What does this PR do?

dependencies are currently installed and listed in an unassociated way with the provider they come from.

From a user POV, associating deps with the specific provider introducing them is crucial in order to eliminate unwanted dependences if using a larger distro like starter.

For example, If I am using starter on a daily basis but then go llama stack build --distro starter... and get 15 NEW dependencies, there is currently NO way for me to tell which provider introduced them.

This work fixes the output of the build process and --print-deps-only

Test Plan

llama stack build --distro starter --image-type venv

Installing normal dependencies for provider 'remote::fireworks'
Using Python 3.12.11 environment at: venv
Resolved 97 packages in 10ms
Uninstalled 3 packages in 48ms
Installed 3 packages in 7ms
 - openai==1.99.9
 + openai==1.78.1
 - protobuf==6.32.0
 + protobuf==5.29.3
 - rich==13.5.3
 + rich==14.1.0
 ...

llama stack build --distro starter --print-deps-only --image type venv

Please install needed dependencies using the following commands:
# Normal Dependencies for remote::cerebras
uv pip install cerebras_cloud_sdk
# Normal Dependencies for remote::ollama
uv pip install ollama aiohttp h11>=0.16.0
# Normal Dependencies for remote::vllm
uv pip install openai
# Normal Dependencies for remote::tgi
uv pip install huggingface_hub aiohttp
# Normal Dependencies for remote::fireworks
uv pip install fireworks-ai
# Normal Dependencies for remote::together
uv pip install together
# Normal Dependencies for remote::bedrock
uv pip install boto3
# Normal Dependencies for remote::nvidia
uv pip install openai
# Normal Dependencies for remote::openai
uv pip install litellm
# Normal Dependencies for remote::anthropic
uv pip install litellm
# Normal Dependencies for remote::gemini
uv pip install litellm
# Normal Dependencies for remote::vertexai

[ashwinb]

@cdoern all the tests are failing here