Auto-reconcile duplicate relationship types in LLM-generated schemas

[NathalieCharbel]

Description

This PR is to guarantee a GraphSchema auto-generated from text by an LLM (SchemaFromTextExtractor) never ends up with two or more relationship_types entries that share the same label.

Rather than hard failing an auto-generated schema, this PR reconciles duplicate relationship types:

• validate_extraction_dict_to_graph_schema now merges same-label relationship_types into a single entry whose properties are the union of the duplicates' properties (de-duplicated by name, first-occurrence wins on conflicts), keeps the first entry's non-property fields, and logs a warning for each merged label. The merge runs on the raw extraction dict before the cross-reference filters and model_validate, so the unioned property set is visible to both - a previously dropped/misreported constraint now validates correctly.

PS:

• why merging instead of splitting: splitting one label into distinct per-pattern types can't be done reliably - nothing binds a relationship-type definition to a specific pattern, so it would require fragile heuristics. Merging is deterministic and keeps the type global per name as Neo4j requires.
• This PR on PR#535 that makes user-provided schemas hard-fail on duplicate relationship types; this PR handles the LLM-generated path by auto-reconciling instead of erroring. The two are independent and compose cleanly.

Type of Change

• New feature
• Bug fix
• Breaking change
• Documentation update
• Project configuration change

Complexity

Complexity: low

How Has This Been Tested?

• Unit tests
• E2E tests
• Manual tests

Checklist

The following requirements should have been met (depending on the changes in the branch):

• Documentation has been updated
• Unit tests have been updated
• E2E tests have been updated
• Examples have been updated
• New files have copyright header
• CLA (https://neo4j.com/developer/cla/) has been signed
• CHANGELOG.md updated if appropriate