Problem (CADETProcess.problem)

Problem is the general problem description, independent of any optimizer: a ParameterSpace (input domain), a MetricSpace (output declarations), and an EvaluationBackend that computes the declared metrics for a named parameter assignment. It is directly usable for sensitivity analysis, design-space exploration, surrogate training, and emulation; OptimizationProblem extends it with objective/constraint policy.

Discipline rule: if a sampler or surrogate would not call it, it does not go on Problem. Direction-aware evaluation, bad_metrics substitution, and callbacks belong to OptimizationProblem.

Problem([parameter_space, metric_space, ...])	General problem description: parameter space, metric space, backend.
EvaluationBackend(*args, **kwargs)	Protocol for anything that evaluates a named parameter assignment.