Create a step-by-step workflow to implement and optimize a large MoE model inference using Colibrì engine. Include: 1) Hardware requirements assessment, 2) Model selection from supported families, 3) Memory hierarchy configuration, 4) Performance benchmarking setup, 5) Real-time monitoring implementation with the web dashboard. Provide code snippets for key configuration steps and performance measurement.