Feature: LangGraph store CE memory connector

## Component
OpenLIT (client)

## What feature would you like to see?
Add a **Community Edition (CE)** memory connector for **LangGraph / LangMem persistent store** so OpenLIT can read and write agent memories held in LangGraph's checkpoint/store backends.

Implementation follows `src/client/src/lib/platform/connectors/memory/README.md`:
- `memory/langgraph/adapter.ts` extending `BaseMemoryAdapter`
- Support common deployment modes (LangGraph Platform API, self-hosted store HTTP API, Postgres-backed store where a query API is available)
- `MemoryTypeDescriptor` describing connection URL, auth, namespace/thread identifiers
- Register in `memory/bootstrap.ts`

**Target capabilities (MVP)**
- `list`, `search`, `get` for stored memory items
- `add`, `update`, `delete` where the store API supports writes
- Filter fields for `thread_id`, `user_id`, `assistant_id`, or LangGraph namespace keys
- Health check

## Why do you need this?
LangGraph is the dominant framework for stateful LLM agents. Teams persist conversation memory and agent state in LangGraph stores but have no way to inspect or manage that memory from OpenLIT's Memory page or Otter chat tools alongside traces.

A CE LangGraph memory connector connects agent observability (traces) with agent memory (store) in one OSS platform — a strong fit for OpenLIT's agent engineering positioning.

## Additional context
- **Edition:** CE (`plan: free`)
- **Open question for implementer:** prioritize LangGraph Platform hosted API vs self-hosted Postgres store REST layer — document supported modes in the adapter descriptor
- **Acceptance criteria:**
- [ ] At least list + search against a test LangGraph store
- [ ] Otter `list_memories` / `search_memories` tools work with connector_id
- [ ] Descriptor-driven form; capability-gated write tools
- [ ] Tests + messages in `en.ts`
- [ ] Connector icon in `public/images/connectors/`

## Pre-submission checklist
- [x] I searched existing issues and didn't find a duplicate

Start with src/client/src/lib/platform/connectors/memory/README.md, then inspect the existing memory adapters and bootstrap.ts to determine the connector contract. Decide which LangGraph deployment mode the MVP supports, implement the descriptor and listed capabilities in memory/langgraph/adapter.ts, and add registration, tests, en.ts messages, and the connector icon. Done means list and search work against a test store, Otter tools accept connector_id, and writes are capability-gated.

Written by the indexing model from the issue text.