docs: correct P-384 build-failure state — v0.6.0 breaks the wrapper staging step

[JC-000]

Summary

Recovers the second commit intended for #64 (it was pushed after #64 had already been merged, so it never landed). Content unchanged from the commit described in #64's PR body.

CLAUDE.md's three P-384 references still described the v0.3.0-era failure (unresolved ec_base384_x/y at link). Verified 2026-07-26 at the v0.6.0 pin: both make p384-overlay and make BACKEND=uci USE_OVERLAY_P384_EMBED=1 now fail earlier, at the ar65 staging step in tools/integration/build_nistcurves_p384.sh (nistcurves_p384_staging/curve/ecdsa384.o never produced — upstream layout drift under the v0.5.0/v0.6.0 bumps). The SHA-384 LUT overlay overflow and the ec_base384_x/y bug likely still lurk behind it.

• Crypto ABI paragraph: no longer claims the P-384 primitives are buildable as an overlay image
• Known issues: describes the current failure chain and records the closure of stale issues make p384-overlay fails to link (unresolved ec_base384_x/y) #32/P-384 TLS handshake regression: first encrypted record (EncryptedExtensions) fails decrypt on hw, blocks wall-clock measurement #45
• P-384 wall-clock section: marked unmeasurable until the build is fixed (the e2e script would just boot the P-256 PRG)

Test plan

• Both P-384 build targets exercised at HEAD 2026-07-26 to confirm the documented failure mode
• Cherry-picked cleanly onto current master (1c0ebaf)

🤖 Generated with Claude Code