Merge upstream/main (#2710 one test binary per crate, #2713) into round3a-integration

Conflicts:
- xtask/src/main.rs: upstream's runners (cargo test or nextest, no
  examples, doctests apart, one wasm `cargo check --keep-going`) with
  #2302's `ci --quick` on top: test_selected builds the same targets
  (--lib --bins --tests), the wasm check takes the changed crates in one
  call, the full test step is upstream's.
- apps/lightcraft/src/control_server.rs: upstream's `before` closure, with
  the batch's no-delay sockets kept.
- docs/parity-checklist.md: LR-BEHAV-ACCESS from main (tests/it path),
  LR-BEHAV-LOCALE / LR-BEHAV-LOCALIZE from the batch (Polish).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

crates/gpu/src/ctx.rs

@@ -118,7 +118,9 @@ impl Drop for Buf {
         let mut b = Some(b);
         let _ = RETIRED.try_with(|r| r.0.borrow_mut().extend(b.take()));
         match b {
-            Some(b) => std::mem::forget(b),
+            // Leaking is deliberate during thread teardown — dropping a wgpu buffer whose device
+            // thread-locals have already gone can abort the process. The device owns it.
+            Some(b) => std::mem::forget(b), // NOSONAR
             None => {
                 RETIRED_BYTES.fetch_add(size, Ordering::Relaxed);
             }

crates/gpu/tests/it/gpu_probe.rs

@@ -1,7 +1,7 @@
 //! What GPU this machine offers the develop pipeline, as one JSON line: the capability probe of
 //! `cargo xtask ci`, which is std + serde_json and has no wgpu of its own. Ignored in normal runs;
 //! xtask runs it after the workspace tests, so it costs no extra build:
-//! `cargo test --workspace --test gpu_probe -- --ignored --nocapture`.
+//! `cargo test --workspace --test it -- --ignored --exact gpu_probe::gpu_probe --nocapture`.
 //!
 //! - `LIGHTCRAFT_GPU_PROBE` unset: the compute device the pipeline picks (`available()`,
 //!   `adapter_name()`, `unavailable_reason()`), the backends it may use and the backends this build
@@ -11,7 +11,6 @@
 //!   a driver that crashes while it is enumerated (issue #136) costs only its own line.
 //!
 //! Only reports: which adapters the pipeline accepts is `ctx.rs`'s policy.
-#![cfg(not(target_arch = "wasm32"))]
 
 use lightcraft_gpu::backend::{self, Backends};
 

crates/gpu/tests/it/main.rs

@@ -0,0 +1,9 @@
+//! The integration tests, one module per file, linked into one test binary.
+//! Binaries of their own: `tests/shutdown.rs` and `tests/shutdown_first.rs` close the GPU for the whole
+//! process, and `tests/memory.rs` checks the process's device-memory counters, which renders on other test
+//! threads would change.
+
+mod equivalence;
+mod fallback;
+#[cfg(not(target_arch = "wasm32"))]
+mod gpu_probe;