Fix issue where vital signs were sometimes being set to NaN

[eedrummer]

Synthea's simulation of height and weight under the age of 20 depends on growth charts published by CDC. The growth charts are provided with a method for transforming values into percentiles and vice-versa. It appears that for higher percentiles, this method can break down.

The case that was used for testing this issue is transforming the ~99.74th percentile BMI for females at 240 months into an actual BMI value.

This code change addresses the issue by converting any request above the 99.5th percentile to just the 99.5th percentile.

This addresses #1372.

[codecov]

Codecov Report

Merging #1381 (34ddcb1) into master (6192ed7) will decrease coverage by 1%.
Report is 5 commits behind head on master.
The diff coverage is 100%.

@@           Coverage Diff            @@
##             master   #1381   +/-   ##
========================================
- Coverage        77%     77%   -1%     
+ Complexity     3685    3681    -4     
========================================
  Files           170     170           
  Lines         24005   24008    +3     
  Branches       3331    3332    +1     
========================================
- Hits          18524   18518    -6     
- Misses         4448    4451    +3     
- Partials       1033    1039    +6     

Files	Coverage Δ
...mitre/synthea/world/concepts/GrowthChartEntry.java	77% <100%> (+8%)	⬆️

... and 27 files with indirect coverage changes

📣 Codecov offers a browser extension for seamless coverage viewing on GitHub. Try it in Chrome or Firefox today!

[dehall]

Code change looks good once the function actually uses the new constant.
Do you have a sense of whether this really only affects the ~1/200 records that are > 99.5 percentile? Or in practice is it less, or more? I'm having a little trouble following the BMI <-> percentile logic as the flow seems to convert one to the other often, meanwhile the height and weight are both also changing.

[dehall]

I ran 200 patients against master and this branch and saw only 31 differences in the observations.csv (out of a total of 137,000 lines). Most were actually "Head Occipital-frontal circumference" but maybe that's just how the dice rolled for that population. Overall looks good