Refactor bioavailability handling

[kylebaron]

Summary

This PR adds another member and methods to the datarecord class to track bioavailability (Fn).

A while ago, we had two different types of records: observations and events and mrgsolve employed some polymorphism to get these to work together.

a936b7e#diff-0e466cbf322abb6ca2747f70501b19721fced39bb696dfb1921ee40067bb5ece

In that setup, Fn was tracked in the event record object.

Since then, we collapsed this down into a single record type and tracking all data in the object (including dose amounts for observation records).

Part of that was to let Fn float around a bit, getting read from $MAIN when a record was getting processed and used in the various functions.

See #1129 for an example of when this doesn't work. So this PR is moving Fn back into the object so that this value will follow the object.

A bunch of tests were added too, including a key test for the problem that motivated this refactor.

Details

• Fn is now a member of datarecord
• Added fn() and fn(double) - getter and setter
• Removed Fn that was getting passed into schedule() and steady(); this was never used
• Data members were re-ordered to put members with same type next to each other; this might be more space efficient
• One test in test-z-alag-f.R was specifically related to Incorrect F when F changes over the course of a lag time #1129; a bunch of other tests added to confirm behavior; a lot of this will also be confirmed in the NONMEM benchmarks

To consider

• Because we had the polymorphism in the code previously, we were working with records as shared pointers. This isn't technically necessary now, but going to keep the code as is because (1) we might want to eventually go back to different record types and (2) there might be some advantage to using shared pointers for these records.

[kylebaron]

This constructor is used when copying data record objects; adding fn as a reminder that this is a key item that needs to make it into the new object.

[kylebaron]

Infusion duration; no longer needs Fn to be passed.

[kylebaron]

These got re-arranged to get all the doubles and bools next to each other.

[kylebaron]

This is key ... if this is a lagged dose (a phantom record) we cannot change bioavailability here.

[kyleam]

I'm jumping ahead of the review request, but my understanding is that this is ready to go (with the already reviewed gh-1131 coming in on top). The background and motivation for moving bioavailability into the data record makes sense to me, and the code change and extensive tests look nicely done.