fix(agent-bot): stabilize webhook delivery for transient upstream failures (#13521)

This fixes the agent-bot webhook delivery path so transient upstream
failures follow the expected delivery lifecycle. Existing fallback
behavior is preserved, and fallback actions are applied only after
delivery attempts are exhausted.

To reproduce, configure an agent-bot webhook endpoint to return 429/500
for message events. Before this fix, failure handling could be applied
too early; after this fix, delivery attempts complete first and then
existing fallback handling runs.

Tested with:
- bundle exec rspec spec/jobs/agent_bots/webhook_job_spec.rb
spec/lib/webhooks/trigger_spec.rb
- bundle exec rubocop spec/jobs/agent_bots/webhook_job_spec.rb
spec/lib/webhooks/trigger_spec.rb

---------

Co-authored-by: Muhsin Keloth <muhsinkeramam@gmail.com>

app/jobs/agent_bots/webhook_job.rb

@@ -1,7 +1,14 @@
 class AgentBots::WebhookJob < WebhookJob
   queue_as :high
+  retry_on RestClient::TooManyRequests, RestClient::InternalServerError, wait: 3.seconds, attempts: 3 do |job, error|
+    url, payload, webhook_type = job.arguments
+    Webhooks::Trigger.new(url, payload, webhook_type || :agent_bot_webhook).handle_failure(error)
+  end
 
   def perform(url, payload, webhook_type = :agent_bot_webhook)
     super(url, payload, webhook_type)
+  rescue RestClient::TooManyRequests, RestClient::InternalServerError => e
+    Rails.logger.warn("[AgentBots::WebhookJob] attempt #{executions} failed #{e.class.name}")
+    raise
   end
 end