Release pipeline: add aarch64 parity

The PR validation workflow builds and tests on aarch64, but the release pipeline (`CreateRelease.yml`) is still x86_64-only. Items to address:

- [ ] **Guest builds**: `build-guests` passes `arch: X64` only — add an `arm64` build and upload artifacts with arch-qualified names
- [ ] **Benchmarks**: `dep_benchmarks.yml` hardcodes `X64` in its `runs-on` label and the benchmark matrix in `CreateRelease.yml` only covers `{hyperv-ws2025, mshv3, kvm} x {amd, intel}` — add aarch64/apple combos (requires runner pool availability)
- [ ] **C API artifacts**: `tar-headers` and `tar-static-lib` in the Justfile hardcode `x86_64-hyperlight-none` paths — add aarch64 equivalents and upload them in the release step
- [ ] **Host build verification**: `build-rust-ubuntu` runs on X64 self-hosted runners only — add an aarch64 build job
- [ ] **Release upload step**: the `gh release create` commands list only x86_64 benchmark and C API tarballs — add the new aarch64 artifacts

Some of these may be blocked on aarch64 self-hosted runner pool availability.

Start with CreateRelease.yml, dep_benchmarks.yml, and the Justfile entries tar-headers and tar-static-lib; trace the existing X64 guest, benchmark, host-build, C API, and release-upload paths. Check aarch64 self-hosted runner availability before implementing the matrix and job changes. Done means aarch64 guest, benchmark, host-build, and C API artifacts are produced, named by architecture, and included in the release upload.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.