Fix export filter keeping old entries when there is an active entry of the same type

[dehall]

Fixes #1383

This PR addresses an issue where the export filter isn't filtering out old entries when there is a separate active entry of the same "type", ie, code. This is especially likely to happen with "Medication Review Due" from the Med Rec module, but it could also happen for any potentially-recurring entry such as Viral Sinusitis or chronic medications.

The issue is that the logic is using record.conditionActive which only checks whether the given code is in the "active" set, it doesn't look at the specific entry at all. The change here is to only look at dates, and the date logic can be pretty simple: "if the entry is open (stop = 0) or the stop is after the cutoff, keep the entry"

[codecov]

Codecov Report

Merging #1385 (574152d) into master (6192ed7) will increase coverage by 0%.
The diff coverage is 80%.

@@           Coverage Diff            @@
##             master   #1385   +/-   ##
========================================
  Coverage        77%     77%           
- Complexity     3685    3704   +19     
========================================
  Files           170     170           
  Lines         24005   23997    -8     
  Branches       3331    3331           
========================================
+ Hits          18524   18587   +63     
+ Misses         4448    4387   -61     
+ Partials       1033    1023   -10     

Files	Coverage Δ
...c/main/java/org/mitre/synthea/export/Exporter.java	62% <80%> (-1%)	⬇️

... and 21 files with indirect coverage changes

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

[jawalonoski]

I'm not sure whether or not to approve this PR. On the one hand, it massively reduces the amount of bloat in the record due to repetition.... mostly related to entries like Medication review due (situation) and employment status...

But it also removes more valuable entries like Normal Pregnancy -- which is double-edged sword. We lose the ability to track how inaccurate our pregnancy model is (the reproduction rate is too high), but instead of fixing it in the model it just "looks better" because they get filtered in the output.

[dehall]

Hmm that's an interesting point. The thing is though we're already filtering those entries out today if there's not an active instance of the same type. This PR is just making the logic more consistent with how we intended it.
For the pregnancy stats, I would say we shouldn't be looking at filtered records anyway when we're reviewing data.