add `pydantic` validators capable of handling `numpyro.distributions` objects

imagine a basic pydantic class such as

from pydantic import BaseModel, NonNegativeFloat
from numpyro.distributions import Distribution
from dynode import DeterministicParameter
from jax.typing import ArrayLike

class Strain(BaseModel):
     r0: Union[NonNegativeFloat, ArrayLike, Distribution, DeterministicParameter]

this class takes static values (NonNegativeFloat), it takes linked deterministic values (DeterministicParameter), it takes prior distributions (Distribution), as well as sampled jax values from said prior distribution (ArrayLike).

Currently there is no validators checking that the support of the Distribution object is actually non-negative. Thus we can pass a prior such as Normal(0, 1) which will be accepted and fit on, however occasionally produce negative R0s and likely break our ODEs. Because of the just-in-time nature of jax, we cannot easily validate the value after it has been sampled (BooleanConversionError), so it is best to validate the distributions themselves.

To mark this issue as complete please add custom distribution types such as NonNegativeDistribution or PositiveDistribution to dynode.typing, similar to dynode.typing.DynodeName

I would recommend building simple greater-than and less-than functions, then combining them together, this will allow us to build more complex validators in the future with the building blocks.

Start with src/dynode/typing/typing.py and inspect DynodeName, then trace how the Strain example's Union fields are validated against numpyro.distributions objects. Define the requested custom distribution types and their greater-than or less-than support checks. Done means invalid distribution supports are rejected while valid distributions remain usable in the shown model.