docs: document Windows and LiteLLM provider setup

[KairosMarco]

Summary

This PR adds a small quickstart clarification for local PageIndex setup:

• recommend a Python virtual environment before installing dependencies,
• add Windows PowerShell activation commands,
• document the common Activate.ps1 execution-policy workaround using only the current shell session,
• show how to use a non-OpenAI LiteLLM provider by setting the provider key and passing --model.

Why

New users can currently follow the README successfully on Unix-like shells, but Windows users may hit PowerShell's script execution policy when activating .venv.

The README also mentions multi-LLM support through LiteLLM, but the quickstart only shows OPENAI_API_KEY. Adding a provider-key example makes it clearer that users can run PageIndex with other LiteLLM-supported providers without changing code.

Changes

• Extend the install step with python3 -m venv .venv and activation.
• Add equivalent Windows PowerShell commands.
• Add Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass as a current-session workaround for blocked virtualenv activation.
• Add a non-OpenAI provider example using DASHSCOPE_API_KEY and --model dashscope/qwen-plus.
• Point users back to LiteLLM provider docs for exact model prefixes and environment variables.

Validation

git diff --check

Local result:

No whitespace errors.