Safer treatment of parameters out of boundaries

[mhliu0001]

Sometimes the parameters proposed by the MCMC are too extreme and fall outside the allowed bounds. In the current implementation, the simulation is still executed in this case. This behavior both wastes computing resources and risks breaking the simulation, since the code is JIT-compiled and does not validate whether invalid parameters are passed to operations such as jnp.random.binomial.

This PR proposes to skip the simulation step entirely when the proposed parameters are out of bounds. In addition, it issues a warning when the number of simulated events in a given bin is zero, so that users can address the issue more promptly—for example, by increasing the batch_size or by imposing more restrictive bounds on the parameters.

[coveralls]

Pull Request Test Coverage Report for Build 21007291517

Details

• 9 of 9 (100.0%) changed or added relevant lines in 2 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage increased (+0.004%) to 84.122%

---

Files with Coverage Reduction	New Missed Lines	%
appletree/likelihood.py	1	85.76%

Totals
Change from base Build 18637300865:	0.004%
Covered Lines:	2437
Relevant Lines:	2897

---

💛 - Coveralls