ROCm-Backend: added and tested docker image functionality with ROCm

[siavashhub]

Summary

• Expands the install-selector tip for Vulkan to include a full ready-to-copy ROCm docker run
  command with the required --device=/dev/kfd, --device=/dev/dri, and --group-add video flags.
• Adds an INFO log line in llamacpp_server.cpp that reports the selected backend at load time,
  making it easier to verify backend selection from logs.
• Updates the LEMONADE_LLAMACPP_BACKEND environment variable to LEMONADE_LLAMACPP throughout
  documentation and Docker examples to match the recent changes.
• Adds a dedicated "Docker Run with CPU backend" section to DOCKER_GUIDE.md.

Test plan

• Verified Docker image starts correctly with LEMONADE_LLAMACPP=rocm on an AMD GPU host
• Confirmed GPU utilization via rocm-smi during inference
• Confirmed LEMONADE_LLAMACPP=cpu and LEMONADE_LLAMACPP=vulkan still work as expected
• Checked log output shows "Using LlamaCpp Backend: rocm/vulkan" on model load
• Reviewed install-selector rendering for the Vulkan GPU path in a browser