Fix Maxwell-Boltzmann Velocity function

[Ellipse0934]

Replacing by the Normal distribution function in Distributions.jl for performance and accuracy improvement.

[Ellipse0934]

function maxwellboltzmann(mass::Real, T::Real) # Currently in Molly
    norm_dist = sum(rand(12)) - 6
    return abs(norm_dist) * sqrt(T / mass)
end
function NEWmaxwellboltzmann(mass::Real, t::Real) # Changed function
    return rand(Normal(0,sqrt(8.3144598*t/mass)))
end

Benchmarks: using @benchmark in BenchmarkTools

BenchmarkTools.Trial:  # Old Function 
  memory estimate:  176 bytes
  allocs estimate:  1
  --------------
  minimum time:     49.129 ns (0.00% GC)
  median time:      50.660 ns (0.00% GC)
  mean time:        61.214 ns (15.26% GC)
  maximum time:     44.944 μs (99.81% GC)
  --------------
  samples:          10000
  evals/sample:     991

BenchmarkTools.Trial: # New Function 
  memory estimate:  0 bytes
  allocs estimate:  0
  --------------
  minimum time:     4.364 ns (0.00% GC)
  median time:      4.789 ns (0.00% GC)
  mean time:        4.806 ns (0.00% GC)
  maximum time:     8.139 ns (0.00% GC)
  --------------
  samples:          10000
  evals/sample:     1000

Also, I think the current function's implementation might be wrong

return abs(norm_dist) * sqrt(T / mass)

By taking the absolute value of the normal distribution the resultant of the normalization of the three velocities does not end up in a Maxwell Distribution. (I began with studying MD with no Chemistry background 2 days ago so please forgive me if I made a huge blunder!)

Blue: Current function and Green: Chi(3) scaled by a factor of sqrt(RT/M)
I also plotted against my function which seems to agree with the scaled Chi(3) and various other sources on the Internet.

[jgreener64]

Great work, thanks for fixing this and improving the implementation.