Allowing broadcast operations to HDF5Dataset

[scottstanie]

A few times I've wanted to do something like this:

f = h5open("file.h5")
ds = f["array"]
newdata = ds .+ 5

But currently that errors:

julia> ds .+ 5
ERROR: MethodError: no method matching iterate(::HDF5Dataset)
Closest candidates are:
  iterate(::Core.SimpleVector) at essentials.jl:600
  iterate(::Core.SimpleVector, ::Any) at essentials.jl:600
  iterate(::ExponentialBackOff) at error.jl:218
  ...
Stacktrace:
 [1] copyto!(::Array{Int64,1}, ::HDF5Dataset) at ./abstractarray.jl:722
 [2] _collect(::UnitRange{Int64}, ::HDF5Dataset, ::Base.HasEltype, ::Base.HasLength) at ./array.jl:564
 [3] collect(::HDF5Dataset) at ./array.jl:558
 [4] broadcastable(::HDF5Dataset) at ./broadcast.jl:664
 [5] broadcasted(::Function, ::HDF5Dataset, ::Int64) at ./broadcast.jl:1234
 [6] top-level scope at REPL[7]:1

(note: the reason it says we need iterate is because of this line in "broadcast.jl":
https://github.com/JuliaLang/julia/blob/master/base/broadcast.jl#L662 that just passes along anything it is not sure of

I have this working for myself by adding:

import Base: broadcastable
broadcastable(x::HDF5Dataset) = read(x)

Is this desirable as a high level feature? or is it too opaque to users to have a dataset automatically get read in? If yes, I can create a pull request to add them here.

Similarly, if we want this I can add the other methods I've included for myself:

collect(x::HDF5Dataset) = collect(read(x))
iterate(x::HDF5Dataset) = iterate(read(x))
iterate(x::HDF5Dataset, state) = iterate(read(x), state)