Prompts exceed LLM token limit

[lauraschauer]

If the references of an advisory are too long, prospector crashes because the LLM token limit gets exceeded.

In some cases, a prompt sent to the LLM might exceed its token limit. For example, in the following cases:

1. When inferring the repository URL: an advisory has a very long description, or (too) many references attached to it.
2. When using the commit classification rule: a diff is too long.

In SAP models, the _call method returns an HTTPError, which must be caught in the function in llm_service.py that invokes the _call method. Currently, this isn't implemented for getting the repository URL.