refactor(ingestion): give the pipeline package a single async entry point

Why:
- The package exposed 8 modules directly, pushing source-type dispatch and
  the ADR-0017 thread-offload obligation onto every caller.

Changes:
- Add parse_and_chunk_document as the sole public entry point.
- Demote the individual parsers to internal/test-only.

tests/unit/application/test_pipeline.py

@@ -0,0 +1,62 @@
+"""Source-type dispatch and thread-offload for `parse_and_chunk_document`."""
+
+import uuid
+
+import pytest
+from anyio import CapacityLimiter
+
+from src.application.ingestion import UnsupportedSourceTypeError, parse_and_chunk_document
+from src.config import ChunkingSettings
+from tests.support.documents import PROSE_DOCX, load_document
+
+pytestmark = [pytest.mark.unit, pytest.mark.asyncio]
+
+
+@pytest.fixture
+def settings() -> ChunkingSettings:
+    return ChunkingSettings()
+
+
+@pytest.fixture
+def limiter() -> CapacityLimiter:
+    return CapacityLimiter(2)
+
+
+async def test_parse_and_chunk_document_docx_returns_chunks(
+    settings: ChunkingSettings, limiter: CapacityLimiter
+) -> None:
+    chunks = await parse_and_chunk_document(
+        load_document(PROSE_DOCX),
+        source_type="docx",
+        file_id=uuid.uuid4(),
+        settings=settings,
+        limiter=limiter,
+    )
+
+    assert chunks
+    assert all(chunk.chunk_index == index for index, chunk in enumerate(chunks))
+
+
+async def test_parse_and_chunk_document_csv_returns_chunks(
+    settings: ChunkingSettings, limiter: CapacityLimiter
+) -> None:
+    data = b"name,value\nfirst,1\nsecond,2\n"
+
+    chunks = await parse_and_chunk_document(
+        data, source_type="csv", file_id=uuid.uuid4(), settings=settings, limiter=limiter
+    )
+
+    assert chunks
+
+
+async def test_parse_and_chunk_document_unsupported_type_raises(
+    settings: ChunkingSettings, limiter: CapacityLimiter
+) -> None:
+    with pytest.raises(UnsupportedSourceTypeError):
+        await parse_and_chunk_document(
+            b"whatever",
+            source_type="doc",
+            file_id=uuid.uuid4(),
+            settings=settings,
+            limiter=limiter,
+        )