fix: count non-dict items in outputs_count

[christian-byrne]

Fix outputs_count not counting non-dict output items (e.g., text strings from PreviewAny node).

Problem

The get_outputs_summary() function in comfy_execution/jobs.py increments count after the isinstance(item, dict) check, causing non-dict items to be skipped from the count.

Text outputs from nodes like PreviewAny come as inline strings ["test"] rather than dicts, so they were never counted. This caused outputs_count: 0 in the jobs API response even when text outputs existed, which prevented the frontend from fetching and displaying them. Example problem:

curl "http://localhost:8188/api/jobs/9c1e37e9-7644-4189-a456-88935b629d70" | jq
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  2243  100  2243    0     0  2589k      0 --:--:-- --:--:-- --:--:-- 2190k
{
  "id": "9c1e37e9-7644-4189-a456-88935b629d70",
  "status": "completed",
  "priority": 0,
  "create_time": 1769675441110,
  "execution_start_time": 1769675441111,
  "execution_end_time": 1769675441129,
  "outputs_count": 0,
  "workflow_id": "d2a78457-e57e-448b-a5dc-b1c4edbf74a0",
  "outputs": {
    "37": {
      "text": [
        "test"
      ]
    }
  },

The solution in this PR aligns OSS Python with Cloud's Go implementation which uses len(itemsArray) to count ALL items regardless of type.

[christian-byrne]

This only affects /jobs API so the risk is low (only consumed by pre-release versions of the frontend). It matches cloud implementation, so it has also already been tested.

[Kosinkadink]

looks good, merging