Support Advanced Numeric Data Distributions (Normal, Exponential, Log-Normal)

[Sauvikn98]

Currently, when our engine generates randomized numeric fields, it defaults to a uniform distribution. While fine for basic mock data, many real-world scenarios require non-uniform probability distributions—such as generating realistic user ages, dynamic pricing, or clustered activity timestamps.

This issue aims to introduce support for Normal, Exponential, and Log-Normal distributions for numeric fields during test data generation.

Tasks:

• Update the TestDataConfig or the numeric SchemaField.constraints type definition to allow specifying a distribution property (e.g., "normal" | "exponential" | "log-normal") along with any necessary math parameters (mean, standard deviation, lambda).
• Modify the data generator logic (likely within TestDataGeneratorService or our Faker utility wrappers) to calculate random values adhering to the provided distribution using our deterministic seedrandom instance.
• Ensure the calculated distributions still gracefully respect existing field boundaries like min and max.
• Add Vitest unit tests verifying that a large generation batch roughly approximates the expected bell curve or decay.

Where to start looking:

• Check the types in TestDataConfig and SchemaField.
• Look at how seedrandom is currently utilized for data generation to ensure your new math functions remain deterministic.