fix: replay channel state on SPA reconnect (#61)

## Summary

When closing and reopening the SPA, channel tabs were not restored because the client relied on localStorage to remember joined channels and re-sent JOIN commands on reconnect. This was fragile and caused spurious JOIN broadcasts to other channel members.

## Changes

### Server (`internal/handlers/api.go`, `internal/handlers/auth.go`)

- **`replayChannelState()`** — new method that enqueues synthetic JOIN messages plus join-numerics (332 TOPIC, 353 NAMES, 366 ENDOFNAMES) for every channel the session belongs to, targeted only at the specified client (no broadcast to other users).
- **`HandleState`** — accepts `?replay=1` query parameter to trigger channel state replay when the SPA reconnects.
- **`handleLogin`** — also calls `replayChannelState` after password-based login, since `LoginUser` creates a new client for an existing session.

### SPA (`web/src/app.jsx`, `web/dist/app.js`)

- On resume, calls `/state?replay=1` instead of `/state` so the server enqueues channel state into the message queue.
- `processMessage` now creates channel tabs when receiving a JOIN where `msg.from` matches the current nick (handles both live joins and replayed joins on reconnect).
- `onLogin` no longer re-sends JOIN commands for saved channels on resume — the server handles it via the replay mechanism, avoiding spurious JOIN broadcasts.

## How It Works

1. SPA loads, finds saved token in localStorage
2. Calls `GET /api/v1/state?replay=1` — server validates token and enqueues synthetic JOIN + TOPIC + NAMES for all session channels into the client's queue
3. `onLogin(nick, true)` sets `loggedIn = true` and requests MOTD (no re-JOIN needed)
4. Poll loop starts, picks up replayed channel messages
5. `processMessage` handles the JOIN messages, creating tabs and refreshing members/topics naturally

closes #60

Co-authored-by: user <user@Mac.lan guest wan>
Co-authored-by: clawbot <clawbot@noreply.git.eeqj.de>
Co-authored-by: Jeffrey Paul <sneak@noreply.example.org>
Reviewed-on: #61
Co-authored-by: clawbot <clawbot@noreply.example.org>
Co-committed-by: clawbot <clawbot@noreply.example.org>

REPO_POLICIES.md

@@ -1,6 +1,6 @@
 ---
 title: Repository Policies
-last_modified: 2026-02-22
+last_modified: 2026-03-09
 ---
 
 This document covers repository structure, tooling, and workflow standards. Code
@@ -98,6 +98,13 @@ style conventions are in separate documents:
   `https://git.eeqj.de/sneak/prompts/raw/branch/main/.gitignore` when setting up
   a new repo.
 
+- **No build artifacts in version control.** Code-derived data (compiled
+  bundles, minified output, generated assets) must never be committed to the
+  repository if it can be avoided. The build process (e.g. Dockerfile, Makefile)
+  should generate these at build time. Notable exception: Go protobuf generated
+  files (`.pb.go`) ARE committed because repos need to work with `go get`, which
+  downloads code but does not execute code generation.
+
 - Never use `git add -A` or `git add .`. Always stage files explicitly by name.
 
 - Never force-push to `main`.
@@ -144,8 +151,14 @@ style conventions are in separate documents:
 - Use SemVer.
 
 - Database migrations live in `internal/db/migrations/` and must be embedded in
-  the binary. Pre-1.0.0: modify existing migrations (no installed base assumed).
-  Post-1.0.0: add new migration files.
+  the binary.
+    - `000_migration.sql` — contains ONLY the creation of the migrations
+      tracking table itself. Nothing else.
+    - `001_schema.sql` — the full application schema.
+    - **Pre-1.0.0:** never add additional migration files (002, 003, etc.).
+      There is no installed base to migrate. Edit `001_schema.sql` directly.
+    - **Post-1.0.0:** add new numbered migration files for each schema change.
+      Never edit existing migrations after release.
 
 - All repos should have an `.editorconfig` enforcing the project's indentation
   settings.