perf: small native tls optimisation

[henderkes]

This folds a 3 instruction tls load on aarch64 to a single mrs.
On x86-64 gnu ld should have relaxed it to a single fs anyway, but this guarantees it.

[Copilot]

Pull request overview

This PR improves runtime performance by forcing a more efficient TLS (thread-local storage) access pattern in the native C code and conditionally enabling -mtls-size=12 when the toolchain supports it, to help compilers/linkers emit shorter TLS sequences (notably on AArch64).

Changes:

• Add a small toolchain probe (mtls-cflags.sh) that prints -mtls-size=12 when compilation+linking succeeds with local-exec TLS.
• Apply the detected TLS CFLAGS in both developer builds (go.sh) and static builds (build-static.sh).
• Force local-exec TLS for key TLS variables in frankenphp.c, and update the performance docs wording around num_threads.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
mtls-cflags.sh	New probe script to conditionally output -mtls-size=12 based on toolchain support.
go.sh	Appends the probe’s output to CGO_CFLAGS for regular Go builds.
frankenphp.c	Introduces THREAD_LOCAL with tls_model("local-exec") on ELF/GNU toolchains and applies it to TLS variables.
docs/performance.md	Updates guidance to configure num_threads via the global frankenphp directive.
build-static.sh	Exports CGO_CFLAGS with the probe result for static build workflows.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.