Implement missing posterior_optima and likelihood_optima for a test model.

[yebai]

Fix some test failures in TuringLang/Turing.jl#2018. Also, transfer some utility functions for test models from Turing to this repo.

[coveralls]

Pull Request Test Coverage Report for Build 5612537845

• 0 of 32 (0.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.9%) to 76.302%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/test_utils.jl	0	32	0.0%

Totals
Change from base Build 5611139343:	-0.9%
Covered Lines:	2051
Relevant Lines:	2688

---

💛 - Coveralls

[codecov]

Codecov Report

Patch coverage has no change and project coverage change: -0.92 ⚠️

Comparison is base (74849d4) 77.22% compared to head (4c81032) 76.30%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #503      +/-   ##
==========================================
- Coverage   77.22%   76.30%   -0.92%     
==========================================
  Files          22       22              
  Lines        2656     2688      +32     
==========================================
  Hits         2051     2051              
- Misses        605      637      +32     

Impacted Files	Coverage Δ
src/test_utils.jl	83.19% <0.00%> (-8.20%)	⬇️

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[sunxd3]

TuringLang/Turing.jl#2049 is failing because the functions were moved here
(c.f. https://github.com/TuringLang/Turing.jl/blob/e32bb71accb4d71dfc5f0377984a00aa5d643c3a/test/modes/OptimInterface.jl#L160 and https://github.com/TuringLang/Turing.jl/blob/e32bb71accb4d71dfc5f0377984a00aa5d643c3a/test/modes/OptimInterface.jl#L191)

These are easy to fix, my concern is that we never refer to Optim interface in DynamicPPL. In my opinion, it is debatable if we should separate the Optim interface and the ground truth to test the interface.

If moving the functions here while leaving everything else put is the way to go , I am happy to see this and TuringLang/Turing.jl#2049 through

[yebai]

These are easy to fix, my concern is that we never refer to Optim interface in DynamicPPL. In my opinion, it is debatable if we should separate the Optim interface and the ground truth to test the interface.

I don't have a strong preference here, but the majority of test-related functionalities for such models are in DynamicPPL. So it might be better to keep all these codes in the same place. In fact, one bug was caused by missing the posterior_optima for newly added models. It shows potential confusion for developers.