Is your feature request related to a problem? Please describe.

Paper: https://arxiv.org/pdf/[2401.15817](https://arxiv.org/pdf/2401.15817)

We should support this in PyRIT!

Describe the solution you'd like

The most straightforward solution I can imagine is a converter that takes in a (benign) image in its constructor and a different (harmful) image in its convert_async method. In that method, it makes the harmful image appear to be benign to evade content filters.