rebinding: populate UTF-8 reason string on SPDM error paths

[sgrams]

Summary

The non-SPDM rebinding_old_prepare / rebinding_new_prepare populate the VMM ReportStatus reason buffer (data: &mut Vec<u8>) on transport failures, gated by #[cfg(feature = "vmcall-raw")]. The SPDM-featured counterparts had the parameter spelled _data and never wrote to it, so GHCI 1.5 ReportStatus (p. 47–48, Tables 3-55/3-56) carried no diagnostic payload when an SPDM rebind failed.

Change

Mirror the non-SPDM pattern at the three SPDM error sites in both rebinding_old_prepare and rebinding_new_prepare:

• spdm_requester / spdm_responder transport setup failure
• with_timeout timeout around spdm_requester_rebind_old / spdm_responder_rebind_new
• the SPDM rebind operation itself

Each .map_err closure now pushes a one-line UTF-8 reason string (operation name, MigrationResult variant, migration ID in hex) into data, gated by #[cfg(feature = "vmcall-raw")] to match the non-SPDM RA-TLS arms. log::error! calls are preserved verbatim.

No behavior change with vmcall-raw off — data is unused, same warning the existing non-SPDM functions already emit on that configuration.

Test

cargo check -p migtd --no-default-features \
  --features main,stack-guard,spdm_attestation,virtio-vsock,policy_v2,test_disable_ra_and_accept_all \
  --target x86_64-unknown-none
cargo check -p migtd --no-default-features \
  --features main,stack-guard,virtio-vsock,policy_v2,test_disable_ra_and_accept_all \
  --target x86_64-unknown-none

Both clean. The vmcall-raw + spdm_attestation build has 3 pre-existing errors on main (mig_socket_info field, setup_transport arity) that are out of scope for this PR.

Reference

GHCI 1.5 April 2026: Tables 3-55 / 3-56 — ReportStatus data buffer carries a UTF-8 reason string describing why a migration/rebind failed.

[haitaohuang]

LGTM