Add JSON file import support to task arguments dialog

[Mbeaulne]

Description

Added JSON file import support to the task arguments dialog alongside existing CSV import functionality. Users can now import both CSV and JSON files to populate pipeline input values. The JSON import supports single objects for direct value mapping and arrays of objects for bulk input mode. Non-string values (objects, arrays, numbers) are automatically JSON.stringified for proper handling.

Enhanced the bulk value parser to be structure-aware, preventing incorrect splitting on commas inside JSON objects, arrays, or quoted strings during bulk input processing.

Related Issue and Pull requests

Type of Change

• Bug fix
• New feature
• Improvement
• Cleanup/Refactor
• Breaking change
• Documentation update

Checklist

• I have tested this does not break current pipelines / runs functionality
• I have tested the changes on staging

Screenshots (if applicable)

Test Instructions

JSON used for testing

[
  {
    "experiment_key": "12345",
    "predictions": "1234",
    "config": { "dslName": "experiment_a", "lr": 0.01, "epochs": 100 },
    "tags": ["production", "v2"],
    "metadata": { "owner": "alice", "team": "ml-platform", "priority": 1 },
    "script": "import torch\nmodel = torch.load('checkpoint.pt')\nresults = model.evaluate(dataset='train')"
  },
  {
    "experiment_key": "1",
    "predictions": "4",
    "config": { "dslName": "experiment_b", "lr": 0.001, "epochs": 200 },
    "tags": ["staging", "v2"],
    "metadata": { "owner": "bob", "team": "ml-platform", "priority": 2 },
    "script": "from sklearn.ensemble import RandomForestClassifier\nclf = RandomForestClassifier(n_estimators=100, max_depth=5)\nclf.fit(X_train, y_train)"
  },
  {
    "experiment_key": "2",
    "predictions": "6",
    "config": { "dslName": "experiment_c", "lr": 0.01, "epochs": 100, "scheduler": { "type": "cosine", "warmup_steps": 500 } },
    "tags": ["production", "v3"],
    "metadata": { "owner": "alice", "team": "recommendations", "priority": 1 },
    "script": "params = {\"batch_size\": 32, \"dropout\": 0.1}\nrun_training(**params)"
  },
  {
    "experiment_key": "9",
    "predictions": "5",
    "config": { "dslName": "experiment_d", "lr": 0.02, "epochs": 50, "scheduler": { "type": "linear", "warmup_steps": 100 } },
    "tags": ["canary"],
    "metadata": { "owner": "carol", "team": "recommendations", "priority": 3 },
    "script": "data = [x for x in range(100) if x % 2 == 0]\nprint(f'filtered {len(data)} items')"
  }
]

1. Navigate to a pipeline task submission dialog
2. Test CSV import functionality to ensure existing behavior is preserved
3. Test JSON import with a single object: {"input1": "value1", "input2": "value2"}
4. Test JSON import with an array for bulk mode: [{"input1": "a"}, {"input1": "b"}]
5. Test JSON import with complex values: {"config": {"lr": 0.01, "epochs": 100}}
6. Verify bulk input parsing handles JSON objects correctly without splitting on internal commas
7. Test error handling with invalid JSON files

Additional Comments

The import button now accepts both .csv and .json files and dynamically handles the appropriate parsing logic based on file extension. The UI messaging adapts to show whether CSV or JSON data was imported, including appropriate terminology for ignored columns/keys.

[github-actions]

🎩 To tophat this PR:

You can add the following URL parameter to your browser to tophat this PR:

`?tophat_location=03-04-bulk_argument_upload_-_json/f1cd6c2`

---

[Mbeaulne]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• Refactor file extension extraction and consolidate test helpers #1896
• Add drag-and-drop file import and multi-format template export to OasisSubmitter #1895
• Add YAML file import support to task arguments dialog #1894
• Add JSON file import support to task arguments dialog #1892 👈 (View in Graphite)
• Add CSV import/export functionality to task arguments dialog #1890
• Add bulk submission support to OasisSubmitter with comma-separated input values #1889
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.