fix pareto_smooth in case of constant tail + cleanup

[avehtari]

pareto_smooth() could fail if x is non-constant, but the tail is constant

Main fix

• Inside pareto_smooth.default(), if ps_tail() returns k as NA, exit early with return(pareto_diags_na(x, return_k, extra_diags))

In addition

• ps_min_ss() now returns NA if k is NA instead of erroring (with the main fix we don't reach this)
• cleaned ps_tail()
  • exit earlier if ndraws_tail < 5
  • check constant tail earlier and return original x without a warning
  • removed duplicate sort.int() call
  • removed redundant constant tail check (which also was not using is_constant())
  • removed min_tail variable as it was not used anymore
• added test to check that pareto_smooth() with non-constant x with constant tail does not error and returns the original x

Copyright and Licensing

By submitting this pull request, the copyright holder is agreeing to
license the submitted work under the following licenses:

• Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
• Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 9db116a is merged into master:

• ✔️as_draws_array: 146ms -> 144ms [-2.81%, +0.33%]
• ✔️as_draws_df: 75.9ms -> 76.4ms [-1.08%, +2.42%]
• ✔️as_draws_list: 170ms -> 171ms [-0.46%, +2.06%]
• ✔️as_draws_matrix: 31.2ms -> 31.5ms [-0.13%, +2.4%]
• 🚀as_draws_rvars: 130ms -> 128ms [-2.97%, -0.23%]
• ❗🐌summarise_draws_100_variables: 764ms -> 801ms [+4.14%, +5.76%]
• ✔️summarise_draws_10_variables: 89.8ms -> 89.2ms [-1.66%, +0.35%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[florence-bockting]

Looks good to me. I just added three additional tests for checking expected behavior of modified parts (i.e., ps_tail and ps_min_ss).

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 8ba5a9f is merged into master:

• ✔️as_draws_array: 144ms -> 144ms [-0.91%, +0.92%]
• ✔️as_draws_df: 76.9ms -> 76.3ms [-2.33%, +0.55%]
• ✔️as_draws_list: 173ms -> 174ms [-0.67%, +1.02%]
• ✔️as_draws_matrix: 31.1ms -> 31ms [-1.4%, +0.57%]
• ✔️as_draws_rvars: 130ms -> 129ms [-1.62%, +0.14%]
• ❗🐌summarise_draws_100_variables: 743ms -> 781ms [+4.46%, +5.98%]
• ✔️summarise_draws_10_variables: 89ms -> 89ms [-1.09%, +1.15%]
  Further explanation regarding interpretation and methodology can be found in the documentation.