feat(server): GgufMetadata reader + SHA-256 sidecar for /props schema-4

[easel]

What

Adds a richer GGUF identity reader on top of Howard Su's existing gguf_inspect (PR #305):

• GgufMetadata struct — captures general.* + <arch>.* header fields (architecture, name, file_type, quantization_version, block_count, embedding_length, context_length, vocab_size) with -1 / "" sentinels distinguishing "not in GGUF" from legitimate zero.
• read_gguf_metadata(path, compute_sha256) — best-effort header read; optional SHA-256 of the whole file.
• Self-contained SHA-256 mini-impl (RFC 6234) — no OpenSSL dependency added for one hash.
• <path>.sha256 sidecar caching — first server start hashes the file (~30s for a 17 GB GGUF on NVMe), subsequent starts read the sidecar. Sidecar I/O failures are non-fatal.
• llama_ftype_name decode — maps general.file_type ints to human-readable names (Q4_K_M, IQ4_XS, etc.) for /props.

Why

/props schema-4 wants a single authoritative "exactly what binary + GGUF + quant + sha256 is loaded" payload so benchmarking and provenance tooling can pin model identity across runs without re-parsing GGUF headers in every consumer. The sidecar makes SHA-256 free after the first boot, which is what makes it usable as a default-on identity field.

Dependencies

None. Purely additive on top of gguf_inspect.{cpp,h} as merged in PR #305 — zero deletions, 333 insertions total. No other server files or build rules change in this PR; /props consumers will be wired up in a follow-up.

Diff stat

 server/src/common/gguf_inspect.cpp | 293 +++++++++++++++++++++++++++++++++++++
 server/src/common/gguf_inspect.h   |  40 +++++
 2 files changed, 333 insertions(+)

Scope / provenance note

This PR is the extracted-and-cleaned remnant of the previously closed PR #336 after a provenance audit. Everything else from that branch (c2_gate, qwen3 drafter changes, structural-defense loaders, and the inadvertent reverts of #273 / #295 / #297) is either landing through its canonical PR (#274) or being dropped entirely. Only this 333-line additive layer on top of Howard's prior gguf_inspect is Erik's own work and survives the audit.

Test plan

• Local build of dflash-server compiles unchanged (no other files touched, so this is mostly a "does it still link" check).
• Manual read_gguf_metadata("…/qwen3-coder-30b-iq4_xs.gguf", true) round-trip on a known model: confirm file_type_name == "IQ4_XS", block_count, context_length match llama-gguf-dump.
• Verify sidecar: delete *.sha256, call once → sidecar appears with 64-hex-char SHA-256 + newline; call again → no re-hash (instrument with a print or strace).
• Verify sentinel behavior on a GGUF missing general.quantization_version: field stays at -1, ok == true.

[cubic-dev-ai]

2 issues found across 2 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>