Split-Shard GGUF & Calibration imatrix.dat Integrity, Alignment & NaN Auditor

Deterministic Multi-shard GGUF 'split.no'/'split.count' cross-header validator + 32/64-byte tensor payload padding checker + llama.cpp imatrix float32 activation-sum variance & NaN auditor engine (.gguf, .dat) built for llama.cpp, ik_llama.cpp, Hugging Face, Bartowski Quant — zero-upload ephemeral Wasm RAM execution.