fix(oban): add executing state to indexing worker unique constraints

Include :executing in the unique states list for both FastIndexingWorker
and MediaCollectionIndexingWorker. This ensures Oban properly tracks
executing jobs and prevents potential duplicate job issues.

Without :executing in the unique states, the uniqueness check doesn't
consider currently running jobs, which could lead to jobs not appearing
in the UI's active tasks view or duplicate jobs being scheduled.

Pinchflat_improvements.md

@@ -159,7 +159,7 @@ docker compose start pinchflat
 3. Log more details when indexing jobs fail
 4. Consider a health check that identifies sources with stale/null `last_indexed_at`
 
-## Issue 7: SlowIndexing jobs don't show as "executing" in UI
+## Issue 7: SlowIndexing jobs don't show as "executing" in UI [COMPLETED]
 
 **Problem:** When `MediaCollectionIndexingWorker` (SlowIndexing) jobs start running, they spawn yt-dlp processes but fail to update their job state from `available` to `executing` in the database. This causes:
 

lib/pinchflat/fast_indexing/fast_indexing_worker.ex

@@ -3,7 +3,7 @@ defmodule Pinchflat.FastIndexing.FastIndexingWorker do
 
   use Oban.Worker,
     queue: :fast_indexing,
-    unique: [period: :infinity, states: [:available, :scheduled, :retryable]],
+    unique: [period: :infinity, states: [:available, :scheduled, :retryable, :executing]],
     tags: ["media_source", "fast_indexing", "show_in_dashboard"]
 
   require Logger

lib/pinchflat/slow_indexing/media_collection_indexing_worker.ex

@@ -3,7 +3,7 @@ defmodule Pinchflat.SlowIndexing.MediaCollectionIndexingWorker do
 
   use Oban.Worker,
     queue: :media_collection_indexing,
-    unique: [period: :infinity, states: [:available, :scheduled, :retryable]],
+    unique: [period: :infinity, states: [:available, :scheduled, :retryable, :executing]],
     tags: ["media_source", "media_collection_indexing", "show_in_dashboard"]
 
   require Logger