[AutoDeploy][chore]: Deprecate and remove transformers mode

🚀 The feature, motivation and pitch

Deprecate and remove the mode="transformers" path in AutoDeploy. The graph mode is now the maintained and default path. The transformers mode was an early lightweight alternative that only applied cached-attention transforms without full-graph capture, but it is no longer actively tested or developed.

Removal scope:

• Remove "transformers" from the mode Literal in LlmArgs
• Remove config/transformers.yaml
• Remove any transformers-mode-specific code paths or branches
• Update documentation referencing transformers mode

Alternatives

Keep transformers mode as a legacy escape hatch, but this adds maintenance burden and test surface with no clear user demand.

Additional context

https://github.com/NVIDIA/TensorRT-LLM/pull/12209 introduces agentic onboarding for models into the graph mode pipeline, thus eliminating the need for a lightweight transformers mode for testing.

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.