False positive missing_const_for_thread_local

Summary

missing_const_for_thread_local is already solved, but it triggers.

See https://github.com/bevyengine/bevy/issues/15241

Lint Name

missing_const_for_thread_local

Reproducer

I tried this code:

#[cfg(feature = "debug_stack")]
thread_local! {
    /// The thread-local [`TypeInfoStack`] used for debugging.
    ///
    /// [`TypeInfoStack`]: crate::type_info_stack::TypeInfoStack
    pub(super) static TYPE_INFO_STACK: std::cell::RefCell<crate::type_info_stack::TypeInfoStack> = const { std::cell::RefCell::new(
        crate::type_info_stack::TypeInfoStack::new()
    ) };
}

I saw this happen:

warning: initializer for `thread_local` value can be made `const`
  --> crates\bevy_reflect\src\serde\de\error_utils.rs:5:1
   |
5  | / thread_local! {
6  | |     /// The thread-local [`TypeInfoStack`] used for debugging.
7  | |     ///
8  | |     /// [`TypeInfoStack`]: crate::type_info_stack::TypeInfoStack
...  |
11 | |     ) };
12 | | }
   | |_^
   |
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#missing_const_for_thread_local
   = note: `#[warn(clippy::missing_const_for_thread_local)]` on by default
   = note: this warning originates in the macro `$crate::thread::local_impl::thread_local_inner` which comes from the expansion of the macro `thread_local` (in Nightly builds, run with -Z macro-backtrace for more info)

I expected to see this happen:
No lint

Version

rustc stable-x86_64-pc-windows-gnu

Additional Labels

No response

Start by reproducing the warning with the provided missing_const_for_thread_local example, then locate the lint implementation and its existing regression tests by searching for the lint name. Trace how the thread_local! expansion is analyzed, and add coverage showing that an initializer already using const produces no warning.