Use combined_score in PromptSampler when available

[junhyeog]

PromptSampler class currently calculates the program's score by averaging all numeric values in the metrics dictionary.
Is there a specific reason for this instead of using the combined_score when it's available in the metrics?
If not, I can submit a PR to use combined_score when it exists.

https://github.com/codelion/openevolve/blob/5d0922250a9560c4e16dd30c138013e28878cd79/openevolve/prompt/sampler.py#L325
https://github.com/codelion/openevolve/blob/5d0922250a9560c4e16dd30c138013e28878cd79/openevolve/prompt/sampler.py#L375
https://github.com/codelion/openevolve/blob/5d0922250a9560c4e16dd30c138013e28878cd79/openevolve/prompt/sampler.py#L440
https://github.com/codelion/openevolve/blob/5d0922250a9560c4e16dd30c138013e28878cd79/openevolve/prompt/sampler.py#L475

[codelion]

Good find, I thought it was fixed but it seems like I only updated the database to use the new get_fitness_score. I will fix it now.