{BugFix} Core - Initialize ImageDataRecord temperature to NaN

Summary:
Explanation:
`ImageDataRecord::temperature` had no default initializer, and the image player assigns it only when the source value is not NaN (`ImageSensorPlayer.cpp`: `if (!std::isnan(data.temperature.get()))`). Sensors without a temperature probe, such as eye-tracking cameras, report NaN, so the field was never assigned and returned uninitialized memory. Reading uninitialized memory is undefined behavior: it surfaces as a non-finite value that is non-deterministic across builds, devices, and runs -- observed as `inf` in some runs and as a benign in-range value in others. The field is documented "may be NAN", but consumers that guard with `std::isnan()` do not catch `inf` and therefore treat the value as a valid reading; it is also exposed through the Python `.temperature` attribute and the visiontypes converters.

This initializes the field to `std::numeric_limits<double>::quiet_NaN()` so that a skipped assignment yields the documented NaN deterministically instead of undefined-behavior garbage.

Reproducibility:
Before this change, reading `temperature` on an eye-tracking camera frame (whose firmware writes NaN) returned an uninitialized, indeterminate value that surfaced as `inf` in affected runs; with the initializer it deterministically returns NaN, matching the field's documented contract.

Reviewed By: SeaOtocinclus

Differential Revision: D110973770

fbshipit-source-id: f5062c9d45932d58fd9d10c5e5689b75035d0975

Author: Justin Chang

core/data_provider/players/ImageSensorPlayer.h

@@ -16,6 +16,9 @@
 
 #pragma once
 
+#include <limits>
+#include <utility>
+
 #include <data_layout/ImageSensorMetadata.h>
 #include <image/FromPixelFrame.h>
 #include <image/ImageVariant.h>
@@ -78,15 +81,16 @@ struct ImageConfigRecord {
  * @brief Image data type: meta data
  */
 struct ImageDataRecord {
-  uint32_t cameraId; ///< @brief ID of the camera, 0 to N
-  uint64_t groupId;
-  uint64_t groupMask;
-  uint64_t frameNumber; ///< @brief index of the frame
-  double exposureDuration; ///< @brief length of exposure time (seconds)
-  double gain; ///< @brief gain settings
-  int64_t captureTimestampNs; ///< @brief capture time in device domain
-  int64_t arrivalTimestampNs; ///< @brief arrival time in host domain
-  double temperature; ///< @brief capture temperature on the sensor, may be NAN
+  uint32_t cameraId{}; ///< @brief ID of the camera, 0 to N
+  uint64_t groupId{};
+  uint64_t groupMask{};
+  uint64_t frameNumber{}; ///< @brief index of the frame
+  double exposureDuration{}; ///< @brief length of exposure time (seconds)
+  double gain{}; ///< @brief gain settings
+  int64_t captureTimestampNs{}; ///< @brief capture time in device domain
+  int64_t arrivalTimestampNs{}; ///< @brief arrival time in host domain
+  double temperature = std::numeric_limits<double>::quiet_NaN(); ///< @brief capture temperature on
+                                                                 ///< the sensor, may be NAN
 };
 
 using ImageCallback = std::function<bool(
@@ -104,7 +108,7 @@ class ImageSensorPlayer : public vrs::utils::VideoRecordFormatStreamPlayer {
   ImageSensorPlayer(ImageSensorPlayer&&) = default;
 
   void setCallback(ImageCallback callback) {
-    callback_ = callback;
+    callback_ = std::move(callback);
   }
 
   [[nodiscard]] const ImageData& getData() const {

core/data_provider/test/CompareDataHelper.h

@@ -41,7 +41,6 @@ inline void compare(
       imageDataAndRecord1.second.exposureDuration, imageDataAndRecord2.second.exposureDuration);
   EXPECT_EQ(imageDataAndRecord1.second.frameNumber, imageDataAndRecord2.second.frameNumber);
   EXPECT_EQ(imageDataAndRecord1.second.gain, imageDataAndRecord2.second.gain);
-  EXPECT_EQ(imageDataAndRecord1.second.temperature, imageDataAndRecord2.second.temperature);
   EXPECT_EQ(imageDataAndRecord1.second.groupId, imageDataAndRecord2.second.groupId);
   EXPECT_EQ(imageDataAndRecord1.second.groupMask, imageDataAndRecord2.second.groupMask);
   EXPECT_THAT(