Registry: Add cloudpickle support

[AgentDosaku]

Summary

Adds a default_materializer argument to Registry / _RegistryCore so users can opt into a fallback serializer (e.g. ZenML CloudpickleMaterializer) when no type-specific materializer is registered.

This enables saving/loading callables and custom classes without first registering explicit materializers for each type.

Closes #309

What changed

• Added default_materializer to:
  • Registry.__init__
  • _RegistryCore.__init__
• Updated materializer resolution order in _find_materializer to include registry-level fallback:
  1. explicitly provided materializer
  2. registered materializer for object type
  3. registered materializer for base classes
  4. registry default_materializer fallback
  5. object self-materializer fallback
• Added load-time fallback behavior for non-importable classes/functions (works with cloudpickle-backed artifacts).
• Added focused unit tests covering the new fallback behavior.

Usage examples (verified)

I ran these examples locally on this branch and confirmed they execute successfully.

from mindtrace.registry import Registry
from zenml.materializers.cloudpickle_materializer import CloudpickleMaterializer

registry = Registry(
    backend="~/.cache/mindtrace/my_registry",
    version_objects=True,
    default_materializer=CloudpickleMaterializer,
)

# Lambda/function serialization
f = lambda x: x + 1
registry.save("test:lambda", f, version="1.0.0")
assert registry.load("test:lambda", version="1.0.0")(1) == 2

from mindtrace.registry import Registry
from zenml.materializers.cloudpickle_materializer import CloudpickleMaterializer

class Point:
    def __init__(self, x):
        self.x = x

    def __reduce__(self):
        return (Point, (self.x,))

registry = Registry(
    backend="~/.cache/mindtrace/my_registry",
    version_objects=True,
    default_materializer=CloudpickleMaterializer,
)

registry.save("test:point", Point(5), version="1.0.0")
point = registry.load("test:point", version="1.0.0")
assert point.x == 5

Testing

• pytest -q tests/unit/mindtrace/registry/core/test_registry.py -k "default_materializer or save_without_materializer"
• ds test --unit (full unit suite on branch)