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/engine/src/files.rs

@@ -2809,7 +2809,7 @@ pub(crate) mod tests {
         let _ = std::fs::remove_dir_all(&dir);
     }
 
-    /// A Sony ARW of `model`: a 64 × 48 uncompressed 14-bit CFA strip in a SubIFD (as `crates/raw/tests/vendor_robust.rs`
+    /// A Sony ARW of `model`: a 64 × 48 uncompressed 14-bit CFA strip in a SubIFD (as `crates/raw/tests/it/vendor_robust.rs`
     /// lays it out) whose IFD carries the full-frame 16–35 mm distortion table (tag 0x7037).
     fn sony_arw(model: &str) -> Vec<u8> {
         use lightcraft_tiff::tags::{self as t, photometric};

crates/engine/src/tests_denoise.rs

@@ -1,7 +1,7 @@
 //! AI denoise through the engine: installing a model, making a photo's denoised picture in the background and for an
 //! export, the Amount slider mixing it in, and the cache. A tiny stand-in model (it halves every sample) takes the
 //! place of a real one, which is enough to see what the machinery does with its answer; how well real models denoise is
-//! measured separately (`crates/denoise/tests/real_model.rs`).
+//! measured separately (`crates/denoise/tests/it/real_model.rs`).
 
 use std::path::{Path, PathBuf};
 use std::sync::Arc;