RXR-373: consistent handling of missing/unrecognized data in calc_ibw()/calc_ffm()/calc_egfr().

[karawoo]

• If required information is missing, we throw an error
• If data is provided but not recognized/supported by the method, we throw a warning and return NULL.
• If data is provided that doesn't match the intended use of the method but it's possible to still calculate the result (e.g. the age is outside the range the method is intended for) we throw a warning but still calculate the value

For calc_egfr() and calc_ffm(), the functions were relatively long with conditional branches for each method. Rather than add more lines I opted to split the code into separate functions for each method, which seemed a bit cleaner, but does make the diff rather large. I added a number of tests for the calc_ffm() methods; calc_egfr() was already relatively well covered so I just added a test for the new behavior there.

[jasmineirx]

this refactor looks great at a quick glance! I'm so glad you tackled this - it had been bothering me for a while.

[jasmineirx]

lgtm - the decision for when to stop versus warn makes sense. I had two comments.

[jasmineirx]

do we want this to be FALSE or NULL?

[karawoo]

Originally we were returning FALSE here so I kept that, but NULL makes sense too. In practice I don't think it should matter too much -- if the egfr method isn't recognized we will have thrown an error long before we get to this line, so I don't think we should ever reach this fallback. Maybe we could remove it altogether.

[jasmineirx]

I know this is how it was written previously, but perhaps this is clearer?

Suggested change

	k <- ifelse(
	preterm & age < 1,
	0.33,
	ifelse(
	age < 1,
	0.45,
	ifelse(
	age > 13 & sex == 'male',
	0.7,
	0.55
	)
	)
	)
	if (age < 1 ) {
	k <- ifelse(preterm, 0.33, 0.45)
	} else if (age > 13) {
	k <- ifelse(sex == 'male', 0.77, 0.55)
	} else {
	k <- 0.55
	}

we then also could only error if sex is not in c("male", "female") if age > 13.