fix: drain HTTP body, paginate Qdrant scroll, deduplicate extractUniqueDirs

[Copilot]

Code review of the VedCode project surfaced three issues.

HTTP connection leak in EnsureCollection

Response body not drained before close, preventing keep-alive connection reuse.

// before
resp.Body.Close()

// after
io.Copy(io.Discard, resp.Body)
resp.Body.Close()

Hardcoded scroll limit of 1000

GetAllFilePoints / GetAllDirPoints used a single scroll request with limit: 1000. Projects exceeding that silently drop records — stale files aren't cleaned up and unchanged files get re-indexed.

Extracted scrollAll() with cursor-based pagination via Qdrant's next_page_offset. Added TestGetAllFilePoints_Pagination covering multi-page retrieval.

Redundant extractUniqueDirs call

indexer.Run() computed extractUniqueDirs(walkResult.Files) twice with the same input. Reused the existing currentDirs variable.

---

📍 Connect Copilot coding agent with Jira, Azure Boards or Linear to delegate work to Copilot in one click without leaving your project management tool.