perf(ci): enable parallel Rust compilation with -Z threads=8

[starpit]

Summary

This PR enables parallel front-end compilation in all GitHub workflows that compile Rust code by adding RUSTFLAGS="-Z threads=8".

Changes

• Added RUSTFLAGS="-Z threads=8" to enable parallel compilation with 8 threads
• Switched all workflows to use Rust nightly toolchain (required for -Z flags)
• Added explanatory comments about the need for nightly and the performance benefit

Updated Workflows

• core.yml - Test and clippy jobs
• release-cli.yml - All build jobs (build-and-upload, build-musl, build-cuda)
• publish_crate.yml - Crate publishing job
• vllm-gce.yml - GCE VM testing job

Performance Impact

The -Z threads=8 flag enables parallel front-end compilation in rustc, which should significantly reduce CI build times, especially for large codebases with many dependencies.

Testing

The changes will be validated by the CI runs on this PR itself.

---

Made with Bob

[starpit]

hmm, nightly seems not to work right now. from core.yml

note: rust-lld: error: undefined symbol: <tokio::util::wake::WakerRef as core::ops::deref::Deref>::deref
          >>> referenced by spnl.44d4a0c7044b1a09-cgu.04
          >>>               /home/runner/work/spnl/spnl/target/debug/deps/spnl-0d0c548af5cdc665.spnl.44d4a0c7044b1a09-cgu.04.rcgu.o:(<tokio::runtime::scheduler::current_thread::CoreGuard>::block_on::<core::pin::Pin<&mut core::pin::Pin<alloc::boxed::Box<core::pin::Pin<&mut dyn core::future::future::Future<Output = core::result::Result<(), anyhow::Error>>>>>>>::{closure#0})
          >>> referenced by spnl.44d4a0c7044b1a09-cgu.04
          >>>               /home/runner/work/spnl/spnl/target/debug/deps/spnl-0d0c548af5cdc665.spnl.44d4a0c7044b1a09-cgu.04.rcgu.o:(<tokio::runtime::scheduler::current_thread::CoreGuard>::block_on::<core::pin::Pin<&mut core::pin::Pin<alloc::boxed::Box<core::pin::Pin<&mut dyn core::future::future::Future<Output = ()>>>>>>::{closure#0})
          >>> referenced by spnl.44d4a0c7044b1a09-cgu.04
          >>>               /home/runner/work/spnl/spnl/target/debug/deps/spnl-0d0c548af5cdc665.spnl.44d4a0c7044b1a09-cgu.04.rcgu.o:(<tokio::runtime::scheduler::current_thread::CoreGuard>::block_on::<core::pin::Pin<&mut core::pin::Pin<&mut dyn core::future::future::Future<Output = core::result::Result<(), anyhow::Error>>>>>::{closure#0})
          >>> referenced 1 more times
          collect2: error: ld returned 1 exit status