Add support to NonMaxSuppressionV5

[lina128]

Tested in local with a python model.

---

This change is 

[pyu10055]

Reviewable status: 0 of 1 approvals obtained (waiting on @lina128 and @pyu10055)

---

tfjs-converter/python/tensorflowjs/op_list/dynamic.json, line 105 at r1 (raw file):

      }
    ],
    "attrs": [

There seems to be an extra attribute 'pad_to_ax_output_size', is it implemented?

---

tfjs-converter/src/executor/model_analysis.ts, line 133 at r1 (raw file):

const CONTROL_FLOW_OPS = ['Switch', 'Merge', 'Enter', 'Exit', 'NextIteration'];
const DYNAMIC_SHAPE_OPS = [
  'NonMaxSuppressionV2', 'NonMaxSuppressionV3', 'NonMaxSuppressionV5', 'Where'

Do we support NMS V4?

[lina128]

Reviewable status: 0 of 1 approvals obtained (waiting on @lina128 and @pyu10055)

tfjs-converter/python/tensorflowjs/op_list/dynamic.json, line 105 at r1 (raw file):

      }
    ],
    "attrs": [

There seems to be an extra attribute 'pad_to_ax_output_size', is it implemented?

tfjs-converter/src/executor/model_analysis.ts, line 133 at r1 (raw file):

const CONTROL_FLOW_OPS = ['Switch', 'Merge', 'Enter', 'Exit', 'NextIteration'];
const DYNAMIC_SHAPE_OPS = [
  'NonMaxSuppressionV2', 'NonMaxSuppressionV3', 'NonMaxSuppressionV5', 'Where'

Do we support NMS V4?

pad_to_ax_output_size is added in V4 and is always set to false in V5. Since we don't support V4 yet, it is safe to not expose it in API. We can start supporting V4, it will be a separate PR.

[pyu10055]

Reviewable status: complete! 1 of 1 approvals obtained