Restrict arguments to Number in comparisons between scalars and EndPoint

[pabloferz]

Looking at some packages invalidations, I found some introduced by this one. As far as I can see, much of these methods aren't supposed to work for things other than Numbers so I think these all should be harmless.

[pabloferz]

I think I spoke too soon, I see that Intervals are expected, at least from the tests, to work with things like Dates. Maybe these really are "irreducible" invalidations, I'll have to think about it a bit more.

[omus]

Currently the only restriction the element type of Interval and AnchoredInterval is that the eltype must be comparable with isless. We may be able to get away with using promotion to convert a scalar (x) into a closed-interval (Interval(x, x)) which could work around the method invalidations by not having to define some of these methods. I suspect there will be a performance hit in doing this though.

[omus]

I'm pretty sure we can update this to be this safely (tests will show if that is the case):

Suggested change

	Base.:(==)(a::Number, b::Endpoint) = a == b.endpoint && isclosed(b)
	Base.:(==)(a::Endpoint, b::Number) = b == a
	Base.:(==)(a::T, b::Endpoint{T}) where T = a == b.endpoint && isclosed(b)
	Base.:(==)(a::Endpoint{T}, b::T) where T = b == a

I'm not sure these are actually used so we may be able to just drop them.

[codecov]

Codecov Report

Merging #143 into master will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #143   +/-   ##
=======================================
  Coverage   75.76%   75.76%           
=======================================
  Files          11       11           
  Lines         491      491           
=======================================
  Hits          372      372           
  Misses        119      119           

Impacted Files	Coverage Δ
src/endpoint.jl	94.59% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c37cbb7...ca64ff6. Read the comment docs.

[pabloferz]

From code coverage, it looks like some of the new definitions are not used in the tests and we might get away with removing them. For others, it might be better to add code tests paths.

I'm not sure how does this look for someone more familiar with direct use cases of the package (I have only use it through other packages that have it as dependency).

[omus]

I'm looking into alternative approaches. I think we may be able to fix Base code such the + and - invalidations no longer occur (see #144). As for converting to a scalar I think we can define an alternative function for this behaviour.

[pabloferz]

That's great. I guess I will remove the arithmetic code changes. Yes, a custom scalar conversion function sounds sensible, but is probably better done in another PR.