Make MuJoCo Python bindings compatible with free-threading.

Introduce a new header `gil.h` defining `MutexLockIfGilDisabled` to support thread-safety in both standard and free-threaded CPython builds.

Protect critical shared states and registries:
- Guard global Python callback pointers in `callbacks.cc` using a mutex. Move `gil_scoped_acquire` into local blocks around refcount modifications to prevent `longjmp` from bypassing destructors.
- Protect raw pointer maps in `structs_wrappers.cc` with static mutexes.
- Replace TOCTOU race in `mjcb_time` initialization with thread-safe `std::call_once`.
- Add synchronization to lazy indexer array cache initialization in `indexers.cc` and `indexer_xmacro.h`.
- Protect vector mutations in `StructListBase::PopulateUpTo` in `structs.h` with a mutex.
- Revert unnecessary atomic changes to threadpool counters.
- Declare free-threading compatibility by passing `pybind11::mod_gil_not_used()` to all extension modules.

Fixes #3259
Fixes #3256
Fixes #2978

PiperOrigin-RevId: 941101502
Change-Id: Iec4ce58afcbc75d4b0be6a9a21fc8a47854242e3

python/mujoco/indexers.h

@@ -15,6 +15,7 @@
 #ifndef MUJOCO_PYTHON_INDEXERS_H_
 #define MUJOCO_PYTHON_INDEXERS_H_
 
+#include <mutex>
 #include <optional>
 #include <string>
 #include <string_view>
@@ -104,6 +105,7 @@ class MjModelGroupedViewsBase {
   std::string name_;
   raw::MjModel* m_;
   pybind11::handle owner_;
+  mutable std::mutex lazy_init_mutex_;
 };
 
 #define XGROUP(MjModelGroupedViews, field, nfield, FIELD_XMACROS) \
@@ -142,6 +144,7 @@ class MjModelIndexer {
   pybind11::handle owner_;
   NameToIDMappings name_to_id_;
   IDToNameMappings id_to_name_;
+  mutable std::mutex lazy_init_mutex_;
 
   // Lazily instantiate a grouped views object when accessed from Python, but
   // cache it once made so that we can return the same one if requested again.
@@ -172,6 +175,7 @@ class MjDataGroupedViewsBase {
   raw::MjData* d_;
   const raw::MjModel* m_;
   pybind11::handle owner_;
+  mutable std::mutex lazy_init_mutex_;
 };
 
 #define XGROUP(MjDataGroupedViews, field, nfield, FIELD_XMACROS) \
@@ -212,6 +216,7 @@ class MjDataIndexer {
   pybind11::handle owner_;
   NameToIDMappings name_to_id_;
   IDToNameMappings id_to_name_;
+  mutable std::mutex lazy_init_mutex_;
 
   // Lazily instantiate a grouped views object when accessed from Python, but
   // cache it once made so that we can return the same one if requested again.