feat: implement Redis Streams support with stream checkpoints and update history

- Added Redis Streams operations to the message bus interface and implementation.
- Introduced StreamCheckpoint model to track last processed stream entry per document.
- Implemented UpsertStreamCheckpoint and GetStreamCheckpoint methods in the Postgres store.
- Created document_update_history table for storing update payloads for recovery and replay.
- Developed update persist worker to handle Redis Stream updates and persist them to Postgres.
- Enhanced Docker Compose configuration for Redis with persistence.
- Updated frontend API to support fetching document state with optional share token.
- Added connection stability monitoring in the Yjs document hook.

backend/scripts/010_add_stream_checkpoints.sql

@@ -0,0 +1,12 @@
+-- Migration: Add stream checkpoints table for Redis Streams durability
+-- This table tracks last processed stream position per document
+
+CREATE TABLE IF NOT EXISTS stream_checkpoints (
+    document_id UUID PRIMARY KEY REFERENCES documents(id) ON DELETE CASCADE,
+    last_stream_id TEXT NOT NULL,
+    last_seq BIGINT NOT NULL DEFAULT 0,
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+CREATE INDEX IF NOT EXISTS idx_stream_checkpoints_updated_at
+    ON stream_checkpoints(updated_at DESC);

backend/scripts/011_add_update_history.sql

@@ -0,0 +1,22 @@
+-- Migration: Add update history table for Redis Stream WAL
+-- This table stores per-update payloads for recovery and replay
+
+CREATE TABLE IF NOT EXISTS document_update_history (
+    id BIGSERIAL PRIMARY KEY,
+    document_id UUID NOT NULL REFERENCES documents(id) ON DELETE CASCADE,
+    stream_id TEXT NOT NULL,
+    seq BIGINT NOT NULL,
+    payload BYTEA NOT NULL,
+    msg_type TEXT,
+    server_id TEXT,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+CREATE UNIQUE INDEX IF NOT EXISTS uniq_update_history_document_stream_id
+    ON document_update_history(document_id, stream_id);
+
+CREATE UNIQUE INDEX IF NOT EXISTS uniq_update_history_document_seq
+    ON document_update_history(document_id, seq);
+
+CREATE INDEX IF NOT EXISTS idx_update_history_document_seq
+    ON document_update_history(document_id, seq);