FINALLY add a benchmark script. We should've had this from the beginning.

[NHDaly]

This is the status as of now:

julia> include("bench/runbench.jl")
  11.000 ns (0 allocations: 0 bytes)
  61.641 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  15.489 ns (0 allocations: 0 bytes)
  4.000 ns (0 allocations: 0 bytes)
  26.104 ns (0 allocations: 0 bytes)
  2.250 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.250 ns (0 allocations: 0 bytes)
  2.250 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
  2.291 ns (0 allocations: 0 bytes)
0

(And here is the above benchmark run on v0.5.1 - when everything used @generated functions. You can see that we have recovered all of that performance.):

julia> include("bench/runbench.jl")
  10.844 ns (0 allocations: 0 bytes)
  62.855 ns (0 allocations: 0 bytes)
  3.041 ns (0 allocations: 0 bytes)
  16.074 ns (0 allocations: 0 bytes)
  4.416 ns (0 allocations: 0 bytes)
  26.230 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.834 ns (0 allocations: 0 bytes)
  2.916 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
  2.833 ns (0 allocations: 0 bytes)
0

[Sacha0]

Thanks Nathan! This is great :).