feat(verifier/snp): support report v3-5, Genoa/Turin and on-demand VCEK from KDS

[jialez0]

Problem

The bare-metal SEV-SNP verifier (deps/verifier/src/snp) currently:

• accepts only report version 2 (report.version != 2 → error), but modern AMD CPUs emit v5 reports;
• bundles only the Milan ASK/ARK/ASVK chain, so Genoa/Turin reports cannot be verified;
• requires the evidence to embed the cert chain (bail!("Cert chain is unset")) and has no KDS fallback, so it fails on hosts that do not provision the VCEK cert table in the extended report.

On a real AMD EPYC 9T25 (Turin) SEV-SNP guest, every genuine attestation is rejected at Cert chain is unset (and would also fail the version / Milan-only checks).

Change

Keeps the sev crate at 4.0.0 (so MSRV 1.76 is preserved — no dependency bump), and:

• accepts report versions 2..=5;
• derives the processor generation (Milan/Genoa/Turin) from the report CPUID family/model bytes;
• bundles Genoa and Turin ASK/ARK/ASVK chains and selects the right one per generation;
• fetches the VCEK from the AMD KDS on demand when the evidence has no cert chain (reuses the crate's existing reqwest dependency);
• reads reported_tcb from the raw report bytes with a generation-aware layout, fixing Turin's FMC-shifted TCB that the sev 4.x struct misreads, and checks the Turin FMC SPL and 8-byte HWID.

Because sev 4.0.0 round-trips a v5 report byte-for-byte, the existing bincode-based signature reconstruction stays valid; only the offset-based fields (CPUID, Turin TCB) are read from the raw bytes.

Validation

Verified end-to-end on a real AMD EPYC 9T25 (Turin) SEV-SNP guest:

• the guest-components SNP attester produces a v5 report with an empty cert chain;
• this verifier now accepts it: VCEK fetched from KDS, ECDSA P-384 signature validated, measurement + reported_tcb (bl=3, tee=2, snp=5, microcode=97, fmc=1) claims emitted.

Adds offline Turin report + VCEK test vectors and 3 unit tests; all 17 snp tests pass, clippy clean, MSRV 1.76 build clean.

Note: the az-snp-vtpm verifier path already references ProcessorGeneration/CERT_CHAINS and is out of scope here; this PR adds ProcessorGeneration which it needs.

[shankailun-aliyun]

@jialez0 ，您好，您的请求已接收，请耐心等待结果。

[shankailun-aliyun]

@jialez0 ，您好，未检测到有镜像需要构建，如需重新检测请评论 /start 。

[shankailun-aliyun]

@jialez0 ，您好，您的请求已接收，请耐心等待结果。

[shankailun-aliyun]

@jialez0 ，您好，未检测到有镜像需要构建，如需重新检测请评论 /start 。

[shankailun-aliyun]

@jialez0 ，您好，您的请求已接收，请耐心等待结果。

[shankailun-aliyun]

@jialez0 ，您好，未检测到有镜像需要构建，如需重新检测请评论 /start 。

[shankailun-aliyun]

@jialez0 ，您好，您的请求已接收，请耐心等待结果。

[shankailun-aliyun]

@jialez0 ，您好，未检测到有镜像需要构建，如需重新检测请评论 /start 。

[shankailun-aliyun]

@jialez0 ，您好，您的请求已接收，请耐心等待结果。

[shankailun-aliyun]

@jialez0 ，您好，未检测到有镜像需要构建，如需重新检测请评论 /start 。