Add @inbounds in alias table sampling

[vancleve]

Seems like one could add @inbounds to this line

StatsBase.jl/src/sampling.jl

Line 624 in 41669cd

x[i] = rand(rng) < ap[j] ? a[j] : a[alias[j]]

In fact its here already when using alias table sampling for Categorical in Distributions:
https://github.com/JuliaStats/Distributions.jl/blob/d6de7214e6d84df77d1eb3ec7a98d23862d1232d/src/samplers/aliastable.jl#L19

[nalimilan]

Sure, why not. Though it seems we need an additional check that the indices of a match those of x.

[nalimilan]

@ParadaCarleton Can you argument on why you want to close this?

[ParadaCarleton]

@ParadaCarleton Can you argument on why you want to close this?

I don't like correctness bugs, and thought this was just hanging around because it was old enough to be before that Yuri post on all the correctness issues in JuliaStats packages.

[LilithHafner]

This optimization would be unsafe because it is possible for make_alias_table! to produce invalid alias tables:

julia> StatsBase.alias_sample!(Random.default_rng(), rand(10), weights(fill(0, 10)), rand(10))
ERROR: BoundsError: attempt to access 10-element Vector{Float64} at index [281471800382896]
Stacktrace:
 [1] throw_boundserror(A::Vector{Float64}, I::Tuple{Int64})
   @ Base ./essentials.jl:14
 [2] getindex
   @ ./essentials.jl:891 [inlined]
 [3] alias_sample!(rng::TaskLocalRNG, a::Vector{Float64}, wv::Weights{Int64, Int64, Vector{Int64}}, x::Vector{Float64})
   @ StatsBase ~/.julia/packages/StatsBase/ebrT3/src/sampling.jl:729
 [4] top-level scope
   @ REPL[10]:1

With @inbounds this would be undefined behavior.