[Bug] Bad encoding for saving optimizations

[ItzAmirreza]

What happened?

I use markdowns and other types of text. That's not the problem until after the optimization, dspy decides to save with non utf-8 encoding. I'm not sure but this might cause different behaviour for loading and reusing that model later on.

Steps to reproduce

from dspy.teleprompt import BootstrapFewShot
optimizer = BootstrapFewShot(metric=metric, max_labeled_demos=4, max_bootstrapped_demos=4)
optimized_program = optimizer.compile(program, trainset=examples)

optimized_program.save("optimized_program.json")

DSPy version

2.6.23