Mark setindex! on SizedArray as possible

[penelopeysm]

StaticArrays.SizedArray is mutable and has setindex! defined on it (method is defined here), so this PR allows BangBang to make use of it:

julia> using BangBang, StaticArrays

julia> v = @SVector zeros(3)
3-element SVector{3, Float64} with indices SOneTo(3):
 0.0
 0.0
 0.0

julia> v2 = similar(v, Real)
3-element SizedVector{3, Real, Vector{Real}} with indices SOneTo(3):
 #undef
 #undef
 #undef

julia> setindex!(v2, 1.0, 1)
3-element SizedVector{3, Real, Vector{Real}} with indices SOneTo(3):
   1.0
 #undef
 #undef

julia> # With this PR
       BangBang.setindex!!(v2, 1.0, 1)
3-element SizedVector{3, Real, Vector{Real}} with indices SOneTo(3):
   1.0
 #undef
 #undef

Right now BangBang.setindex!! errors as it tries to go through an immutable path, which hits JuliaArrays/StaticArrays.jl#1330:

julia> v3 = similar(v, Real)
3-element SizedVector{3, Real, Vector{Real}} with indices SOneTo(3):
 #undef
 #undef
 #undef

julia> BangBang.setindex!!(v3, 1.0, 1)
ERROR: UndefRefError: access to undefined reference
Stacktrace:
 [1] getindex
   @ ./essentials.jl:917 [inlined]
 [2] getindex
   @ ~/.julia/packages/StaticArrays/IZgIP/src/SizedArray.jl:92 [inlined]
 [3] macro expansion
   @ ~/.julia/packages/StaticArrays/IZgIP/src/deque.jl:193 [inlined]
 [4] _setindex
   @ ~/.julia/packages/StaticArrays/IZgIP/src/deque.jl:186 [inlined]
 [5] setindex
   @ ~/.julia/packages/StaticArrays/IZgIP/src/deque.jl:185 [inlined]
 [6] _setindex
   @ ~/ppl/BangBang.jl/ext/BangBangStaticArraysExt.jl:27 [inlined]
 [7] may
   @ ~/ppl/BangBang.jl/src/core.jl:11 [inlined]
 [8] setindex!!(xs::SizedVector{3, Real, Vector{Real}}, v::Float64, I::Int64)
   @ BangBang ~/ppl/BangBang.jl/src/base.jl:478
 [9] top-level scope
   @ REPL[15]:1

[MasonProtter]

Thank you!

[penelopeysm]

Thanks @MasonProtter! I would ask you for a release; but maybe I can do one better.

Turing.jl currently already uses BangBang quite heavily and in the near future that reliance will only increase. Seven of the eight most recent open issues are from current or ex-Turing devs, and I ran into a few more problems getting it to work with DimensionalData which I haven't gotten around to reporting :-)

I am the only one left on the project now, but if you would like, I am happy to help maintain the repo and maybe even try to fix some of them. Obviously PRs will still need to be reviewed, but I can help with the grunt work of releases, CI, and stuff.

[MasonProtter]

Yes, that would be great!

[penelopeysm]

Amazing, thank you! I shall try to register and hope the bot picks up collaborator status...