Ensure inlined_ids eltype of StdVector is correct

[omus]

The GetObjectRefs function requires that a StdVector{ObjectID} is passed in for the inlined_ids argument. If we update serializer.object_ids to return a Set{ObjectID} things break in an unexpected way as StdVector(collect(Set{ObjectID}())) isa StdVector{Any} where as StdVector(collect(Set{ObjectIDAllocated}())) isa StdVector{ObjectID}.

Here's an overview which may help those trying to understand what's going on:

julia> using ray_julia_jll: ObjectID, ObjectIDAllocated

julia> using CxxWrap.StdLib: StdVector

julia> StdVector(ObjectIDAllocated[])
0-element CxxWrap.StdLib.StdVectorAllocated{ObjectID}

julia> StdVector(ObjectID[])
0-element CxxWrap.StdLib.StdVectorAllocated{Any}

julia> StdVector{ObjectID}(ObjectIDAllocated[])
ERROR: MethodError: no method matching StdVector{ObjectID}(::Vector{ObjectIDAllocated})

Closest candidates are:
  StdVector{ObjectID}()
   @ ray_julia_jll ~/.julia/packages/CxxWrap/aXNBY/src/CxxWrap.jl:624

Stacktrace:
 [1] top-level scope
   @ REPL[10]:1

julia> StdVector(ObjectIDAllocated[])::StdVector{ObjectID}
0-element CxxWrap.StdLib.StdVectorAllocated{ObjectID}

julia> StdVector(ObjectID[])::StdVector{ObjectID}
ERROR: TypeError: in typeassert, expected StdVector{ObjectID}, got a value of type CxxWrap.StdLib.StdVectorAllocated{Any}
Stacktrace:
 [1] top-level scope
   @ REPL[12]:1

The type checks here should provide some protection about accidental changes.

Related to: JuliaInterop/CxxWrap.jl#367

[codecov]

Codecov Report

Merging #112 (654d12a) into main (5c1c1fd) will not change coverage.
The diff coverage is 50.00%.

@@           Coverage Diff           @@
##             main     #112   +/-   ##
=======================================
  Coverage   75.06%   75.06%           
=======================================
  Files           8        8           
  Lines         369      369           
=======================================
  Hits          277      277           
  Misses         92       92           

Flag	Coverage Δ
Ray.jl	75.06% <50.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
src/runtime.jl	53.72% <50.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[kleinschmidt]

seems liek the root cause here is that ObjectID is an abstract type adn julia just punts and turns it into Any? Or is ti CxxWrap's fault? Either way LGTM

[kleinschmidt]

orthogonal to this PR but it seems liek these lines are identical to what's below and fiddly enough that they'd benefit from refactoring

[omus]

I've left these lines duplicated for now as the Python code has minor differences. If we adopt those differences in our Julia code they will differ.

[omus]

seems liek the root cause here is that ObjectID is an abstract type adn julia just punts and turns it into Any? Or is ti CxxWrap's fault? Either way LGTM

This seems to be another CxxWrap corner case with empty vectors as having any elements in the Vector is fine:

julia> x = [FromRandom(ObjectID)]
1-element Vector{ObjectIDAllocated}:
 ObjectIDAllocated(Ptr{Nothing} @0x0000600001cd8660)

julia> StdVector(x)
1-element CxxWrap.StdLib.StdVectorAllocated{ObjectID}:
 ray_julia_jll.ObjectIDDereferenced(Ptr{Nothing} @0x0000600001e6c3c0)

julia> y = ObjectID[FromRandom(ObjectID)]
1-element Vector{ObjectID}:
 ObjectIDAllocated(Ptr{Nothing} @0x0000600001cd8630)

julia> StdVector(y)
1-element CxxWrap.StdLib.StdVectorAllocated{ObjectID}:
 ray_julia_jll.ObjectIDDereferenced(Ptr{Nothing} @0x0000600001e8a790)