Incidence Rate uses multiple target cohort episodes per person

[chrisknoll]

Expected behavior

IR should only use the first cohort episode per person.

Actual behavior

IR will use multiple cohort episodes, leading to a result that looks like there were more people used in the analysis than actually in the cohort.

Steps to reproduce behavior

1. Create a target cohort with multiple episodes per person (ie: a drug exposure cohort that has a fixed-offset exit)
2. Create an IR analysis, and specify the TAR as start and end.
3. Observe: person count in the analysis will be the number of episodes in the cohort, not the number of people, because multiple episodes were used: it should restrict to first episode per person.