Add reference to TrainingPeaks blog post with TSS formula

TrainingPeaks' blog has a post by Joe Friel that has the TSS formula,
and it uses normalized power. Add this to the references as my only
other reference for the TSS formula was the 2003 Coggan post, which
says to use average power, even though I knew normalized power is the
correct value to use.

Author: mtraver

Diff for: README.md

@@ -376,6 +376,8 @@ Coggan, Andrew. (2003, March 13). _TSS and IF - at last!_ Retrieved June 14, 201
 
 Eckner, Andreas. (2017, April 3). _Algorithms for Unevenly Spaced Time Series: Moving Averages and Other Rolling Operators._ Retrieved June 14, 2017, from http://eckner.com/papers/Algorithms%20for%20Unevenly%20Spaced%20Time%20Series.pdf
 
+Friel, Joe. (2009, Sept 21). _Estimating Training Stress Score (TSS)._ Retrieved June 22, 2017, from https://www.trainingpeaks.com/blog/estimating-training-stress-score-tss/
+
 # License
 This project is licensed under the MIT License. See
 [LICENSE](https://github.com/mtraver/fitanalysis/blob/master/LICENSE) file

Diff for: fitanalysis/activity.py

@@ -404,11 +404,12 @@ def training_stress(self, ftp):
     """Calculates the training stress of the activity.
 
     This is essentially a power-based version of Banister's heart rate-based
-    TRIMP (training impulse). See (Coggan, 2003) cited in README for details.
-    The formula in that post specifies that average power should be used in
-    the calculation, but normalized power is used here instead because it
-    yields values in line with the numbers from TrainingPeaks; using average
-    power does not.
+    TRIMP (training impulse). Andrew Coggan's introduction of TSS and IF
+    specifies that average power should be used to calculate training stress
+    (Coggan, 2003), but a later post on TrainingPeaks' blog specifies that
+    normalized power should be used (Friel, 2009). Normalized power is used
+    here because it yields values in line with the numbers from TrainingPeaks;
+    using average power does not.
 
     Args:
       ftp: Functional threshold power in Watts.