Allow overriding `PydanticPrompt`'s instruction and examples in the `__init__` method by changing them to instance attributes with fallback

[Glinte]

Describe the Feature
I expect to be able to initialize a subclass of PydanticPrompt and overriding the examples/instruction in the __init__ directly.

i.e.

ContextRecallClassificationPrompt(instruction="...", examples=[...])

Why is the feature important for you?
I am doing some evals and using LLMContextRecall / other related MetrciWithLLM classes, they provide a default prompt which is convenient, but overriding the prompt is not so convenient. For example, ContextRecallClassificationPrompt is the prompt class used in LLMContextRecall, and its __init__ method does not allow changing the examples (because they are class attributes), so the only way to go around it is to subclass ContextRecallClassificationPrompt and change the examples, or to initialize an instance of ContextRecallClassificationPrompt and then setting the instruction/examples (prompt.examples = examples). Both are more verbose than I would expect, and feels like ragas is making things harder than necessary.

[anistark]

Closing this as obsolete.

• New metrics use function-based prompts, not PydanticPrompt classes
• You can customize by modifying prompt functions directly
• The pain point of "overriding class attributes in init" no longer exists