Add Dynamo renderer transport

[AmeenP]

Summary

Adds a renderer transport selector to verifiers so RendererClient can choose between the existing vLLM /inference/v1/generate wire shape and Dynamo's /chat/completions + nvext.token_data wire shape.

• adds ClientConfig.renderer_transport, defaulting to prime_vllm_generate
• forwards the selected transport into renderers.client.generate(...)
• pins the standalone PrimeIntellect-ai/renderers repo to the Dynamo transport PR head (7ca1ab3, renderers Question: Completion IDs for multi-turn conversations. #11)
• adds unit coverage that the renderer client passes the configured transport through

Context

Latest verifiers/main now consumes renderers from the standalone repo, not packages/renderers. This PR uses that latest path and pins the companion renderers PR directly.

Companion PR: PrimeIntellect-ai/renderers#11

Validation

• uv lock --check
• uv run pytest tests/test_renderer_client.py
• uv run ruff check pyproject.toml verifiers/types.py verifiers/clients/renderer_client.py tests/test_renderer_client.py
• uv run ruff format --check pyproject.toml verifiers/types.py verifiers/clients/renderer_client.py tests/test_renderer_client.py