FIX: Make post created/edited automation trigger regexes more specific (#32944)

- For `with_images`, ignore `<img>` tags representing emoji or a quote
avatar
- For `with_links` and `with_uploads`, exclude tags starting with the
letter "a" other than `<a>` (like `<aside>`)
- For `with_code`, the `<pre>` tag wasn't being detected because it was
expecting an extra char before the ">"

plugins/automation/lib/discourse_automation/event_handlers.rb

@@ -102,11 +102,14 @@ module DiscourseAutomation
             cooked = post.cooked
             # note the only 100% correct way is to lean on an actual HTML parser
             # however triggers may pop up during the post creation process, we can not afford a full parse
-            next if post_features.include?("with_images") && !cooked.match?(/<img[^>]+>/i)
-            next if post_features.include?("with_links") && !cooked.match?(/<a[^>]+>/i)
-            next if post_features.include?("with_code") && !cooked.match?(/<pre[^>]+>/i)
+            if post_features.include?("with_images") &&
+                 !cooked.match?(/<img(?![^>]*class=["'](emoji|avatar))[^>]*>/i)
+              next
+            end
+            next if post_features.include?("with_links") && !cooked.match?(/<a\s+[^>]*>/i)
+            next if post_features.include?("with_code") && !cooked.match?(/<pre[^>]*>/i)
             if post_features.include?("with_uploads") &&
-                 !cooked.match?(/<a[^>]+class=["']attachment[^>]+>/i)
+                 !cooked.match?(/<a\s+[^>]*class=["']attachment[^>]*>/i)
               next
             end
           end