Fix sparse broadcast macro hygiene issue

[Sacha0]

Due to a macro hygiene issue introduced in #17302, broadcast methods specialized for unary operations over SparseMatrixCSCs are not being called:

julia> versioninfo()
Julia Version 0.6.0-dev.804
Commit 182f4b3* (2016-09-27 03:28 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin15.5.0)
  CPU: Intel(R) Core(TM) i7-3520M CPU @ 2.90GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Sandybridge)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.7.1 (ORCJIT, ivybridge)

julia> sin.(spdiagm(1.0:5.0))
5×5 Array{Float64,2}:
 0.841471  0.0       0.0       0.0        0.0
 0.0       0.909297  0.0       0.0        0.0
 0.0       0.0       0.14112   0.0        0.0
 0.0       0.0       0.0      -0.756802   0.0
 0.0       0.0       0.0       0.0       -0.958924

This PR fixes that issue,

julia> sin.(spdiagm(1.0:5.0))
5×5 sparse matrix with 5 Float64 nonzero entries:
        [1, 1]  =  0.841471
        [2, 2]  =  0.909297
        [3, 3]  =  0.14112
        [4, 4]  =  -0.756802
        [5, 5]  =  -0.958924

and introduces tests to guard against regression. Best!

[Sacha0]

Absent objections or requests for time, I plan to merge this Saturday (assuming I can work up the first-merge nerve!). Best!

[vtjnash]

Not an objection, but I'm somewhat surprised we want to 'esc'ape the function name.

[Sacha0]

Not an objection, but I'm somewhat surprised we want to 'esc'ape the function name.

Without escaping the function name, the macro hygiene system replaces the function name with a distinct identifier. Illustration:

julia> macroexpand(:(@oldversion parentmeth childmeth))
quote
    #5#broadcast(::typeof(childmeth),#7#A::SparseMatrixCSC) = begin  # REPL[3], line 4:
            parentmeth(childmeth,#7#A)
        end
end

julia> macroexpand(:(@newversion parentmeth childmeth))
quote
    broadcast(::typeof(childmeth),#9#A::SparseMatrixCSC) = begin  # REPL[4], line 5:
            parentmeth(childmeth,#9#A)
        end
end

So the new method does not attach to the correct function (broadcast). Thoughts? Thanks!

[tkelman]

similar-ish to #18122 ?

[Sacha0]

similar-ish to #18122 ?

Perhaps tenuously? :)