fw/services/hrm_manager: fix assert in prv_system_task_hrm_handler

Check if data is available in the circular buffer before attempting to
read in prv_system_task_hrm_handler(). This prevents an assertion
failure when system task callbacks are queued without corresponding
events in the buffer, which can occur during high system load such as
voice dictation session startup.

The race condition occurs when:
- prv_event_put() queues an event and calls system_task_add_callback()
- Multiple callbacks may be queued rapidly during system load
- One callback processes and consumes the event
- Another callback runs but finds the buffer empty

Instead of asserting and crashing, we now log a warning and return
gracefully. HRM functionality continues normally as new events arrive.

Signed-off-by: Joshua Jun <joshuajun@proton.me>

Author: jplexer

src/fw/services/common/hrm/hrm_manager.c

@@ -369,6 +369,19 @@ static void prv_system_task_hrm_handler(void *context) {
   time_t utc_now = rtc_get_time();
 
   mutex_lock_recursive(s_manager_state.lock);
+
+  // Check if there's data available in the circular buffer before attempting to read
+  const uint16_t available_bytes =
+      circular_buffer_get_read_space_remaining(&s_manager_state.system_task_event_buffer);
+  if (available_bytes < sizeof(PebbleHRMEvent)) {
+    // No event available to read - this can happen if system task callbacks are queued
+    // without corresponding events, or during concurrent access.
+    PBL_LOG(LOG_LEVEL_WARNING, "HRM: system task handler called with no event in buffer "
+            "(available=%u, needed=%u)", available_bytes, sizeof(PebbleHRMEvent));
+    mutex_unlock_recursive(s_manager_state.lock);
+    return;
+  }
+
   PebbleHRMEvent event;
   prv_read_event_from_buffer_and_consume(&s_manager_state.system_task_event_buffer,  &event);