add with_* methods

[d-v-b]

adds methods for creating copies of models with new fields

[d-v-b]

from pydantic_zarr.experimental.v3 import ArraySpec, GroupSpec

g = GroupSpec(attributes={"foo": 10}, members={})
a = ArraySpec(
    shape=(10,), 
    attributes={}, 
    data_type='int8',
    codecs=({"name": "bytes"},),
    fill_value=0,
    chunk_key_encoding={"name": "default", "configuration": {"separator": "/"}},
    chunk_grid={"name": "regular", "configuration": {"chunk_shape": (1,)}}
    )

# Copy of g with new members
new_g = g.with_members({"a": a})
print(new_g)
"""
zarr_format=3 attributes={'foo': 10} node_type='group' members={'a': ArraySpec(zarr_format=3, node_type='array', attributes={}, shape=(10,), data_type='int8', chunk_grid={'name': 'regular', 'configuration': {'chunk_shape': (1,)}}, chunk_key_encoding={'name': 'default', 'configuration': {'separator': '/'}}, fill_value=0, codecs=({'name': 'bytes'},), storage_transformers=(), dimension_names=None)}
"""

[d-v-b]

worth noting that because these with_* methods trigger validation, you can't easily do this:

a = ArraySpec(shape=(10, 20), ...)
# this will fail validation against the chunk grid
b = a.with_shape((1,2,3))

but if we offered a method that let you update both the chunk grid AND the shape at the same time, then this pattern would work. with_shape_and_chunk_grid is a bit long and I don't have a better idea right now. This is something for a future PR