feat(rewrite): Add another LLM detection step to augment entities

Priority Level

Medium (Nice to have)

Is your feature request related to a problem?

We'd like the initial entity detection for Rewrite to be very broad. It is only when the Sensitivity Disposition looks at the risk of entities in combination with other entities that the true risk can really be determined. If the entity does not contribute to re-identification risk, then it will choose the disposition of "leave-as-is" and it will not impact the rewrite.

We currently do have the occasional issue of LLM-as-a-Judge for privacy saying too many discriminating attributes are left in the anonymized text. This will help with that.

Describe the solution you'd like

Add one more step in the Rewrite pipeline that uses an LLM to find additional entities. Use domain supplements to guide the prompt.

Describe alternatives you've considered

No response

Additional context

No response

Start by tracing the Rewrite pipeline and its existing entity-detection step, then inspect how domain supplements guide prompts. The work is done when an additional LLM-based detection step augments the entities used by Rewrite while allowing Sensitivity Disposition to decide which entities remain unchanged.