Synthetic Dataset: Support setting a shared prompt prefix.

[sjmonson]

Added support for a prefix_tokens argument which allows for generating a dataset where every prompt contains a shared prefix of length prefix_tokens.

Closes #104

[Copilot]

Pull Request Overview

This PR adds support for a shared prompt prefix by introducing a new configuration option (prefix_tokens) and modifying prompt generation in the synthetic dataset.

• Introduces a new field (prefix_tokens) in the dataset configuration.
• Generates and prepends a shared prefix to each prompt, updating token count calculations accordingly.

Comments suppressed due to low confidence (1)

src/guidellm/dataset/synthetic.py:191

• The function name '_create_prompt' suggests that it returns a prompt string, but it now returns a list of token IDs. Consider renaming it (e.g., '_create_prompt_tokens') or updating its documentation to clarify the returned type.

def _create_prompt(self, prompt_tokens: int, start_index: int) -> list[int]: