Add model 73-77 build targets and runtime registrations

[cortensor-ryuma]

• add new model Dockerfiles and build scripts for model IDs 73-77
• register new CPU/GPU model IDs in llm.py for gemma4:e4b, gemma4:26b, gemma4:31b, qwen3.6:27b, and qwen3.6:35b
• wire the same model IDs so runtime container selection resolves to cts-llm-73 through cts-llm-77
• extend the docker model range cap from 67 to 77
• keep the change confined to model/build registration paths without altering unrelated runtime behavior