An End-to-End Coding Guide to NVIDIA KVPress for Long-Context LLM Inference, KV Cache Compression, and Memory-Efficient Generation