use AppleAccelerate as default on Mac

[vpuri3]

w/o AppleAcceleate

julia> using Metal, BenchmarkTools, LinearAlgebra

julia> N = 512; A = rand(Float32, N, N); B = rand(Float32, N, N); C = rand(Float32, N, N); @btime mul!($C, $A, $B);
  904.208 μs (0 allocations: 0 bytes)

with AppleAcceleate

julia> using Metal, BenchmarkTools, LinearAlgebra, AppleAccelerate

julia> N = 512; A = rand(Float32, N, N); B = rand(Float32, N, N); C = rand(Float32, N, N); @btime mul!($C, $A, $B);
  189.125 μs (0 allocations: 0 bytes)

discussion: https://julialang.slack.com/archives/C689Y34LE/p1691412094284679
cc: @ViralBShah @oscardssmith

[ViralBShah]

#355