improve REPL printing of sets

[rfourquet]

Compared to arrays and dicts, the default REPL printing of sets is less than ideal for inspecting values.
Master:

julia> Set(rand(100))
Set([0.9355287178152043, 0.4928021501163431, 0.30965545692897645, 0.12977832204668327, 0.21199393799696042, 0.0049669103311482665, 0.21544680124723192, 0.94211
31090732746, 0.1757484374656193, 0.9827906624667768  …  0.9395587064783664, 0.6661801938225571, 0.6137788625422396, 0.8401334617771041, 0.1553607216380093, 0.8
201636488737938, 0.7213369975241664, 0.946783923950687, 0.9626916317380967, 0.9239185348013443])

PR:

julia> Set(rand(100))
Set{Float64} with 100 entries:
  0.2570901274036239
  0.07658455754806237
  0.16876087627652803
  0.6875079709294418
  0.4490615526957693
  0.9858242705954456
  0.24446960302165
  0.7804304622569547
  0.3584736990667057
  0.10787858192490529
  0.01588708187652732
  0.6316046571808667
  0.45151298050579336
  0.3619189322408094
  0.2619333768991221
  ⋮

[KristofferC]

It is a bit different from how Array prints the ⋮, cf:

julia> a = rand(100)
100-element Array{Float64,1}:
 0.4746815092073091
 0.9484816361221338
 0.613811448828379
 0.5142341357868621
 0.05873629586930762
 0.702505071916633
 0.45982443303453424
 ⋮
 0.30567035367024586
 0.06633560832376939
 0.7123200471052471
 0.4799413972061428
 0.49309094787857055
 0.5925469710385052
 0.28220057407708987

Not sure if it matters.

[rfourquet]

It is a bit different from how Array prints the ⋮

It's on purpose, there is no particular ordering for sets, so printing the "last" elements (skippling middle ones) is not specially meaningfull compared to just printing elements until there is no more space. Note that this is how dicts are shown at the REPL (ellipsis at the end). Also, I plan to propose changing the show(::Set) method to put the ellipsis at the end.

[KristofferC]

Makes sense.

[JeffBezanson]

I know dicts use "entries" but I think "elements" makes more sense for sets.

[rfourquet]

Good point, I copy-pasted from dicts' method without thinking twice.

[rfourquet]

Given the support received by this PR, I went ahead and fixed test failures. I think this would be good to go on my side if CI turns green.

[rfourquet]

This issue is now tested in other places of the code.

[StefanKarpinski]

I like it and am tempted to merge but could you just add a NEWS note about the change?

[rfourquet]

Good to go? triage?

[rfourquet]

Oh I had not noticed you merged it, thanks!