Add awake detection classifiers and optimize REM detection logic

- Introduced `refine_awake_detection.py` for enhanced awake detection using heart rate samples and sleep stages.
- Added `two_stage_classifier.py` implementing a two-stage classification approach for distinguishing awake vs sleep and NREM vs REM stages.
- Created `two_stage_v2.py` to incorporate consecutive awake signals requirement for improved accuracy.
- Updated `remOptimizedClassifier.ts` to utilize learned awake prior probabilities based on time since sleep onset, enhancing the prediction model.

Author: jeremymanning

.gitignore

@@ -280,3 +280,5 @@ credentials.json
 notes/screenshots/*.png
 hardcoded_findings.txt
 notes/fitbit_merged_training_data.json
+notes/raw_sleep_data.json
+scripts/parse_fitbit_takeout.py