[SPRINT-02-02] Split collection pulling from SBOL document indexing

[Gonza10V]

Parent sprint: #87
Depends on: #88
Recommended order: 2
Codex-ready: yes

Goal

Refactor collection indexing so SynBioHub fetching and SBOL document indexing are separate steps.

Background

Today _index_collections(collections) does too much:

1. assumes every collection is a SynBioHub URI;
2. pulls each URI into self.sbol_doc;
3. scans implementations;
4. classifies strains, plasmids, backbones, enzymes, and ligases;
5. mutates indexed_plasmids, indexed_backbones, restriction_enzyme_implementations, and ligase_implementations;
6. does additional strain and component-definition scans.

This makes it hard to debug whether a failure is due to a fetch problem, local SBOL structure, role mismatch, implementation linking, or indexing logic.

Scope

Extract the existing behavior into smaller helpers. Suggested shape:

def pull_collection_uris(self, uris: list[str]) -> sbol2.Document: ...
def index_sbol_document(self, doc: sbol2.Document, source: str = "local") -> InventoryAudit | dict: ...
def _index_implementation(self, implementation: sbol2.Implementation, doc: sbol2.Document, source: str): ...
def _index_strain_module(self, strain: sbol2.ModuleDefinition, implementation: sbol2.Implementation | None, doc: sbol2.Document): ...
def _index_plasmid_or_backbone_definition(self, definition: sbol2.ComponentDefinition, implementation: sbol2.Implementation | None, doc: sbol2.Document): ...
def _index_reagent_implementation(self, implementation: sbol2.Implementation, built_object): ...

The exact names may differ. The important requirement is separation of concerns.

Non-goals

• Do not change biological classification semantics unless tests expose a bug.
• Do not implement the full future inventory/collection_indexer.py architecture unless it is the smallest clean path.
• Do not remove existing constructor behavior yet.

Acceptance criteria

• Existing SynBioHub URI constructor path still works.
• Local document indexing can call the same SBOL document indexing helper.
• Fetch failures are distinguishable from indexing failures.
• Indexing helpers are individually testable or at least covered through targeted integration tests.
• No hidden PUDU/Opentrons dependency is introduced.

Verification

Run:

pytest -k "index or collection or synbiohub"
ruff check .

Codex implementation notes

• Keep PR size moderate; extract functions without rewriting every caller.
• Preserve current list-based state until a later sprint introduces a richer inventory object.
• Avoid broad renames unless they simplify tests and reduce duplicated code.