Improve HMC sampler setup for faster convergence

[yebai]

Fix #569

We could have obtained better ESS and Rhat via NUTS, but the example is meant to demonstrate HMC. This PR improves the HMC setup to get better convergence.

[github-actions]

Preview the changes: https://turinglang.org/docs/pr-previews/570
Please avoid using the search feature and navigation bar in PR previews!

[penelopeysm]

Previously:

Summary Statistics
  parameters      mean       std      mcse   ess_bulk   ess_tail      rhat   e ⋯
      Symbol   Float64   Float64   Float64    Float64    Float64   Float64     ⋯

          s²    5.5479    1.2616    0.9774     1.8639    17.5367    1.5297     ⋯
           m   -0.0440    0.3888    0.3512     1.3504    10.8871    2.0833     ⋯
        x[1]    2.6569    0.2115    0.0582    13.6021    41.4668    0.9993     ⋯
        x[2]   -1.5160    0.3819    0.2243     3.3737    25.5550    1.2711     ⋯

With this PR:

Summary Statistics
  parameters      mean       std      mcse   ess_bulk   ess_tail      rhat   e ⋯
      Symbol   Float64   Float64   Float64    Float64    Float64   Float64     ⋯

          s²    3.0615    3.6123    0.4927    60.3417    92.1985    1.0231     ⋯
           m   -0.8087    1.5927    0.3190    28.5407    28.0818    1.0590     ⋯
        x[1]   -0.8858    1.9850    0.3445    32.7690    94.9785    1.0091     ⋯
        x[2]   -0.7899    2.1671    0.4259    28.8392    24.6639    1.0233     ⋯