Additional functionality tests for SmithWilson curves

[alecloudenback]

To test/extend API:

Q: Are the curves composable (is + defined), like in the quickstart in README
A: Yes, e.g.:

julia> discount(sw_swq,1)
1.0208956877292932

julia> discount(sw_swq+ Yields.Constant(0.05),1)
0.9713151170101941

Q: How are the yields in the quote instruments defined? Unsure if they are coninuous/periodic compounded.

• Either way, need to document default assumption
• Allow for alternate construction, e.g. with Periodic(0.03,2) (ie bond equivalent yields)

[alecloudenback]

cc @kasperrisager

[kasperrisager]

Q1: I'd say Smith-Wilson curves are composable, and I think they would currently just fall back to AbstractYield behaviour. But the current definition of +/- take place in Periodic(1) rate convention, and that's not very natural for Smith-Wilson, Continuous() would be more natural. Like it would if you add a discounting function and a survival probability function. So I'm really not sure what the best way to do +/- for SW is... Maybe it should be illegal to add two AbstractYields, and instead you have to do AbstractYield + AbstractSpread where the AbstractSpread defines which convention to do the addition in? (That's not a real suggestion, haven't thought it through).

Q2: I implicitly thought that the rate and interest arguments would be Periodic(frequency). That convention is implicit in cashflows. I tend to think this is the only natural choice (so we should just document). Would there be a meaningful quoting convention where this is not the case, like biannual payments quoted in annual compounding?