Make searchsorted on ClosedInterval generic over AbstractVector

[iamed2]

Previously, searchsorted with ClosedInterval as the index could only search a Range.
But, since searchsorted(first|last) works for any type with an ordering, we can extend searchsorted to search any AbstractVector.

This allows forcing Dimensional behaviour with SortedVector the same way one can force Categorical behaviour with CategoricalVector.

julia> A = AxisArray([1,3,2], Axis{:a}(SortedVector(["one", "three", "two"])))
1-dimensional AxisArray{Int64,1,...} with axes:
    :a, String["one", "three", "two"]
And data, a 3-element Array{Int64,1}:
 1
 3
 2

Before:

julia> A["three"]
ERROR: MethodError: no method matching isless(::String, ::IntervalSets.ClosedInterval{String})
Closest candidates are:
  isless(::AbstractString, ::AbstractString) at strings/basic.jl:124
  isless(::IntervalSets.ClosedInterval, ::IntervalSets.ClosedInterval) at /Users/ericdavies/.julia/v0.6/AxisArrays/src/intervals.jl:26
  isless(::Union{Base.Dates.AbstractTime, Number}, ::IntervalSets.ClosedInterval) at /Users/ericdavies/.julia/v0.6/AxisArrays/src/intervals.jl:53
Stacktrace:
 [1] searchsorted(::AxisArrays.SortedVector{String}, ::IntervalSets.ClosedInterval{String}, ::Int64, ::Int64, ::Base.Order.ForwardOrdering) at ./sort.jl:139
 [2] searchsorted(::AxisArrays.SortedVector{String}, ::IntervalSets.ClosedInterval{String}, ::Base.Order.ForwardOrdering) at ./sort.jl:213
 [3] axisindexes(::Type{AxisArrays.Dimensional}, ::AxisArrays.SortedVector{String}, ::String) at /Users/ericdavies/.julia/v0.6/AxisArrays/src/indexing.jl:190
 [4] getindex(::AxisArrays.AxisArray{Int64,1,Array{Int64,1},Tuple{AxisArrays.Axis{:a,AxisArrays.SortedVector{String}}}}, ::String) at /Users/ericdavies/.julia/v0.6/AxisArrays/src/indexing.jl:123

After:

julia> A["three"]
3

julia> A["three".."three"]
1-dimensional AxisArray{Int64,1,...} with axes:
    :a, String["three"]
And data, a 1-element Array{Int64,1}:
 3

julia> A["thra".."thro"]
1-dimensional AxisArray{Int64,1,...} with axes:
    :a, String["three"]
And data, a 1-element Array{Int64,1}:
 3

julia> A["a".."a"]
1-dimensional AxisArray{Int64,1,...} with axes:
    :a, String[]
And data, a 0-element Array{Int64,1}

julia> A["a".."z"]
1-dimensional AxisArray{Int64,1,...} with axes:
    :a, String["one", "three", "two"]
And data, a 3-element Array{Int64,1}:
 1
 3
 2

[nickrobinson251]

Bump. Is this still a thing we want and makes sense? :)