Allow custom ValidationPlugins in iter_validate_collection()

[turbomam]

Collection.iter_validate_collection() hardcodes JsonschemaValidationPlugin. Projects like NMDC have custom ValidationPlugin subclasses (e.g. NmdcSchemaValidationPlugin) that can't be used through linkml-store's validation path.

Proposal: Accept an optional plugins parameter:

def iter_validate_collection(self, plugins=None, **kwargs):
    if plugins is None:
        plugins = [JsonschemaValidationPlugin(closed=True)]
    ...

Note: The MongoDBLoader for linkml validate is being tracked in linkml/linkml#3380 — it belongs in linkml core rather than here, since it's ~15 lines of pymongo and shouldn't require a linkml-store dependency. This issue is about the separate question of making linkml-store's own validation flexible.

Context: microbiomedata/nmdc-schema#2969

[turbomam]

Update — MongoDBLoader moving to linkml core

After discussion, the MongoDBLoader for linkml validate belongs in linkml core (linkml/linkml#3380) as an optional [mongodb] extra. The validator loader is a thin adapter — it only needs to implement iter_instances() with pymongo. Putting it in linkml-store would mean projects that just want linkml validate to read MongoDB would need to depend on the full linkml-store stack, which isn't the right dependency direction for a single-purpose loader.

linkml-store's MongoDB support is a different layer — full CRUD, indexing, search, inference. Both can coexist without overlap.

This issue is still relevant for the separate question of accepting custom ValidationPlugin instances in iter_validate_collection(), per the updated title/body.