Fix/infusion now

[kylebaron]

Summary

Users are allowed to trigger discontinuities in the simulation (e.g. doses) from within the model. This is done through an evdata object which can be scheduled at some time in the future or the user can ask for it to happen now.

There was a bug in this when an infusion was given by this now mechanism: there was no code to turn the infusion off.

As I was addressing this issue, I realized that there are some other expected behaviors for these doses that weren't getting implemented when these now doses were triggered.

Details

Workflow

• I was testing github actions and got a workflow file committed; intent is to remove that where ever it shows up

test-cpp.R

• Unit tests for events triggered from within the model

devtran.cpp

• This is where most of the action is
• prob.rate_main() can be called regardless of whether the record just came from the data or not; whenever the rate attribute is non-negative, this call will do nothing; added logic in rate_main() to make this clear
• Added some logic to do nothing with the modeled event when evid is 0; this is an observation
• After creating the event (new_ev) we have to process it like other event records; get and check the bioavailability, call rate_main(), etc
• Added logic to turn off an infusion

odeproblem.cpp

• Added code to return early if rate is not negative

[kyleam]

Thanks for the summary and details. Some of devtran.cpp changes are a bit beyond my grasp but overall look good to me.

I confirmed that make test-cpp passes on this PR and fails as expected when I revert the code changes.