[Feature] Minimal IUCLID export (#338)

This is an initial implementation that creates a working minimal .i6z document.
It passes schema validation and can be imported into IUCLID.

Caveat:
IUCLID files target individual compounds.
Pathway is not actually covered by the format.

It can be added in either soil or water and soil OECD endpoints.
**I currently only implemented the soil endpoint for all data.**

This sort of works, and I can report all degradation products in a pathway (not a nice view, but we can report many transformation products and add a diagram attachment in the future).

Adding additional information is an absolute pain, as we need to explicitly map each type of information to the relevant OECD field.
I use the XSD scheme for validation, but unfortunately the IUCLID parser is not fully compliant and requires a specific order, etc.

The workflow is: finding the AI structure from the XSD scheme -> make the scheme validation pass -> upload to IUCLID to get obscure error messages -> guess what could be wrong -> repeat 💣

New specifications get released once per year, so we will have to update accordingly.
I believe that this should be a more expensive feature, as it requires significant effort to uphold.

Currently implemented for root compound only in SOIL:

- Soil Texture 2
- Soil Texture 1
- pH value
- Half-life per soil sample / scenario (mapped to disappearance; not sure about that).
- CEC
- Organic Matter (only Carbon)
- Moisture content
- Humidity

<img width="2123" alt="image.png" src="attachments/d29830e1-65ef-4136-8939-1825e0959c62">
<img width="2124" alt="image.png" src="attachments/ac9de2ac-bf68-4ba4-b40b-82f810a9de93">
<img width="2139" alt="image.png" src="attachments/5674c7e6-865e-420e-974a-6b825b331e6c">

Reviewed-on: enviPath/enviPy#338
Co-authored-by: Tobias O <tobias.olenyi@envipath.com>
Co-committed-by: Tobias O <tobias.olenyi@envipath.com>

epiuclid/schemas/loader.py

@@ -0,0 +1,90 @@
+"""Load and cache IUCLID XSD schemas with cross-reference resolution.
+
+The bundled XSD schemas use bare ``schemaLocation`` filenames (e.g.
+``platform-fields.xsd``, ``commonTypesDomainV10.xsd``) that don't match the
+actual directory layout.  This module builds an explicit namespace → file-path
+mapping so that ``xmlschema`` can resolve every import.
+"""
+
+from __future__ import annotations
+
+from functools import lru_cache
+from pathlib import Path
+
+import xmlschema
+
+_SCHEMA_ROOT = Path(__file__).resolve().parent / "v10"
+
+# Namespace → relative file-path (from _SCHEMA_ROOT) for schemas that are
+# referenced by bare filename from subdirectories that don't contain them.
+_NS_LOCATIONS: dict[str, str] = {
+    "http://iuclid6.echa.europa.eu/namespaces/platform-fields/v1": "platform-fields.xsd",
+    "http://iuclid6.echa.europa.eu/namespaces/platform-metadata/v1": "platform-metadata.xsd",
+    "http://iuclid6.echa.europa.eu/namespaces/platform-container/v2": "platform-container-v2.xsd",
+    "http://iuclid6.echa.europa.eu/namespaces/platform-attachment/v1": "platform-attachment.xsd",
+    "http://iuclid6.echa.europa.eu/namespaces/platform-modification-history/v1": (
+        "platform-modification-history.xsd"
+    ),
+    "http://www.w3.org/1999/xlink": "xlink.xsd",
+    "http://www.w3.org/XML/1998/namespace": "xml.xsd",
+    "http://iuclid6.echa.europa.eu/namespaces/domain/v10": ("domain/v10/commonTypesDomainV10.xsd"),
+    "http://iuclid6.echa.europa.eu/namespaces/oecd/v10": ("oecd/v10/commonTypesOecdV10.xsd"),
+}
+
+# doc_type → (subdir, filename-pattern)
+_DOC_TYPE_PATHS: dict[str, tuple[str, str]] = {
+    "SUBSTANCE": ("domain/v10", "SUBSTANCE-10.0.xsd"),
+    "REFERENCE_SUBSTANCE": ("domain/v10", "REFERENCE_SUBSTANCE-10.0.xsd"),
+}
+
+
+def _absolute_locations() -> list[tuple[str, str]]:
+    """Return (namespace, absolute-file-URI) pairs for all known schemas."""
+    return [(ns, (_SCHEMA_ROOT / rel).as_uri()) for ns, rel in _NS_LOCATIONS.items()]
+
+
+def _esr_path(subtype: str) -> Path:
+    """Return the path to an Endpoint Study Record schema."""
+    return _SCHEMA_ROOT / "oecd" / "v10" / f"ENDPOINT_STUDY_RECORD-{subtype}-10.0.xsd"
+
+
+def _doc_type_path(doc_type: str, subtype: str | None = None) -> Path:
+    if doc_type == "ENDPOINT_STUDY_RECORD":
+        if not subtype:
+            raise ValueError("subtype is required for ENDPOINT_STUDY_RECORD schemas")
+        return _esr_path(subtype)
+    info = _DOC_TYPE_PATHS.get(doc_type)
+    if info is None:
+        raise ValueError(f"Unknown document type: {doc_type}")
+    subdir, filename = info
+    return _SCHEMA_ROOT / subdir / filename
+
+
+@lru_cache(maxsize=32)
+def get_content_schema(doc_type: str, subtype: str | None = None) -> xmlschema.XMLSchema:
+    """Return a compiled XSD schema for validating content elements.
+
+    Parameters
+    ----------
+    doc_type:
+        IUCLID document type (``SUBSTANCE``, ``REFERENCE_SUBSTANCE``,
+        ``ENDPOINT_STUDY_RECORD``).
+    subtype:
+        Required for ``ENDPOINT_STUDY_RECORD`` (e.g. ``BiodegradationInSoil``).
+    """
+    path = _doc_type_path(doc_type, subtype)
+    return xmlschema.XMLSchema(str(path), locations=_absolute_locations())
+
+
+@lru_cache(maxsize=1)
+def get_document_schema() -> xmlschema.XMLSchema:
+    """Return a compiled XSD schema for the ``platform-container-v2`` wrapper.
+
+    This validates the full ``<Document>`` element (PlatformMetadata + Content +
+    Attachments + ModificationHistory).  Content is validated with
+    ``processContents="strict"`` via ``xs:any``, but only if the content
+    namespace has been loaded.  For full content validation, use
+    :func:`get_content_schema` separately.
+    """
+    path = _SCHEMA_ROOT / "platform-container-v2.xsd"
+    return xmlschema.XMLSchema(str(path), locations=_absolute_locations())