Add source kinds and the Learn tab with Learning sources (#2)

Every Source now has a kind: News, Learning or Fun. Curated Sources default
to News, and custom Sources persist their kind (ones saved before default
to News). Each kind has its own time window: News follows the 12/24/48 h
setting, Learning reaches back 14 days, Fun 3 days.

A new Learn tab between Briefing and Sources lists Learning articles with
the same paid/sponsored filters and reader view as Today. It is built from
a list of kinds so a Fun section can follow. Today and the Briefing only
use News. ByteByteGo is the first Learning source, on by default.

The home shell names its tabs instead of using hard-coded indexes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

lib/services/feed_service.dart

@@ -23,14 +23,15 @@ class FeedService {
 
   static const _timeout = Duration(seconds: 20);
 
-  /// Fetches all [sources] in parallel. Failing feeds are reported in
+  /// Fetches all [sources] in parallel and drops articles older than the age
+  /// [maxAgeOf] gives for their Source. Failing feeds are reported in
   /// [FeedResult.errors] instead of failing the whole refresh.
   ///
   /// [corsProxy] is only used on web, where browsers block cross-origin feed
   /// requests. It is prepended to the feed URL, e.g. `https://proxy/?url=`.
   Future<FeedResult> fetchAll(
     List<NewsSource> sources, {
-    required Duration maxAge,
+    required Duration Function(NewsSource source) maxAgeOf,
     String corsProxy = '',
   }) async {
     final errors = <String, String>{};
@@ -43,12 +44,15 @@ class FeedService {
       }
     }));
 
-    final cutoff = DateTime.now().toUtc().subtract(maxAge);
+    final now = DateTime.now().toUtc();
     final seen = <String>{};
     final articles = <Article>[
-      for (final list in results)
+      for (final (i, list) in results.indexed)
         for (final a in list)
-          if ((a.published == null || a.published!.isAfter(cutoff)) && seen.add(a.id)) a,
+          if ((a.published == null ||
+                  a.published!.isAfter(now.subtract(maxAgeOf(sources[i])))) &&
+              seen.add(a.id))
+            a,
     ];
     articles.sort((a, b) {
       final pa = a.published, pb = b.published;