Andrew Jesson

This PR builds on the merged GEPA types work to add Python client integration. It enables Python users to create `GEPAConfig` objects and launch GEPA optimization jobs through the TensorZero...

**Problem:** The `Optimizer` trait (tensorzero-optimizers/src/lib.rs:94-106) accepts `val_examples: Option`, making validation examples optional. However, some optimizers like GEPA require validation examples and currently error when None is provided (see tensorzero-optimizers/src/gepa/mod.rs:48-54). **Proposed...

Implements an offline version [GEPA](https://arxiv.org/abs/2507.19457), a prompt optimization algorithm that uses genetic programming to evolve prompt templates through iterative evaluation and mutation. ## Key Files - **Main optimization loop**: `tensorzero-optimizers/src/gepa/lib.rs`...