gguf_hash.py: Add sha256

[mofosyne]

Found out that python hashlib that is already included in this repo has sha256 in addition to the sha1 we are already using for sha1 and uuid hash.

We have already implemented this in the C version of llama-gguf-hash so might as well include it in the python version.

• I have read the contributing guidelines
• Self-reported review complexity:
  • Low
  • Medium
  • High

---

Cross Checking gguf_hash.py with the C implementation llama-gguf-hash

$ ~/llama.cpp/gguf-py/scripts/gguf_hash.py Maykeye_Tinyllama-5M-v0.2-F16.gguf --no-layer
sha1      a9de42f2bbeee1eba49bc39b25cf69ff7a0937f6  Maykeye_Tinyllama-5M-v0.2-F16.gguf
sha256    8b3e00226cc2a55398b1ffbda7af8464040f9cd7b22ccbef8ba60b227924a2b1  Maykeye_Tinyllama-5M-v0.2-F16.gguf
uuid      86b1ebff-d754-50fc-9245-d23fe329817c  Maykeye_Tinyllama-5M-v0.2-F16.gguf

$ ~/llama.cpp/llama-gguf-hash --all --no-layer --uuid Maykeye_Tinyllama-5M-v0.2-F16.gguf
xxh64     cbd383cfd4c897e6  Maykeye_Tinyllama-5M-v0.2-F16.gguf
sha1      a9de42f2bbeee1eba49bc39b25cf69ff7a0937f6  Maykeye_Tinyllama-5M-v0.2-F16.gguf
sha256    8b3e00226cc2a55398b1ffbda7af8464040f9cd7b22ccbef8ba60b227924a2b1  Maykeye_Tinyllama-5M-v0.2-F16.gguf
uuid      86b1ebff-d754-50fc-9245-d23fe329817c  Maykeye_Tinyllama-5M-v0.2-F16.gguf

---

(side thought: Would it be worth having general.sha256 to have self validating gguf files... sounds interesting... but likely only makes sense in the context of cryptographic signing of tensor data... which will require more thought on standardizing the authenticity and verification of tensor weights as not being tampered with... and llama.cpp is a bit too new to do something like that in my opinion)