llama.cpp Release b11601 Adds Qualcomm Hexagon HTP Power Modes

Llama.cpp release b11601 introduces configurable Qualcomm Hexagon HTP DCVS and HMX power modes via a new --hex-power-mode runner flag. This update allows users to fine-tune hardware management and keep the Hexagon Tensor Processor awake during dynamic voltage and frequency scaling profiles. This release enhances edge AI deployment efficiency on Snapdragon devices by giving developers direct control over power and performance profiles for Hexagon NPUs. Better hardware management helps optimize power consumption and execution speed when running large language models on mobile and embedded hardware. The power mode setting is integrated into the unified run.py script and sets the GGML_HEXAGON_POWER_MODE environment variable prior to initializing the Hexagon backend registry. This implementation avoids standard CLI argument parsing because the environment variable must be established before the backend initializes.