feat: add execute/executions CLI commands and refactor generate (#180)

## Summary

- **New `vlmrun execute` command** — runs agent executions via
`/v1/agent/execute` with support for skills (`--skill`, `--skill-id`),
toolsets (`--toolset`), file inputs (`-i`), schema (`--schema`), and
full agent name (`--name <agent>:<version>`). Supports `--wait` with
polling and `--format json` output.
- **New `vlmrun executions list/get` subcommands** — list and retrieve
agent execution results with rich formatted output (panel layout,
colorized status, duration computed from timestamps).
- **Refactored `vlmrun generate`** — replaced separate `image` and
`document` subcommands with a single unified command (`vlmrun generate
-i <file>`) that auto-detects file type from extension and routes to the
appropriate API endpoint (`/v1/document/generate` for images/documents,
`/v1/video/generate`, `/v1/audio/generate`). Supports skills, schema,
batch/wait modes, and JSON output.
- **Improved `vlmrun predictions list`** — formatted output matching
executions style with panel layout, colorized status, and duration
computed from timestamps. Removed N+1 API call overhead (no longer
fetches each prediction individually).
- **Fixed truthiness checks** in `predictions get` and `executions get`
to correctly display `0` values for usage fields (elements_processed,
credits_used, etc.).
- **Updated `CreditUsage` model** with `steps`, `message`, and
`duration_seconds` fields.
- **Version bump** to 0.6.0.

## Test plan

- [ ] `vlmrun execute -i <file> --skill-id <skill> --wait` completes and
displays result
- [ ] `vlmrun executions list` shows formatted table with duration
- [ ] `vlmrun executions get <id>` shows full details including credits
- [ ] `vlmrun generate -i photo.jpg --domain image.caption` auto-detects
image type
- [ ] `vlmrun generate -i doc.pdf --domain document.invoice` routes to
document endpoint
- [ ] `vlmrun predictions list` is fast (no N+1 API calls) and shows
formatted output
- [ ] `vlmrun predictions get <id>` shows 0 values correctly (not
hidden)


Made with [Cursor](https://cursor.com)
<!-- devin-review-badge-begin -->

---

<a href="https://app.devin.ai/review/vlm-run/vlmrun-python-sdk/pull/180"
target="_blank">
  <picture>
<source media="(prefers-color-scheme: dark)"
srcset="https://static.devin.ai/assets/gh-open-in-devin-review-dark.svg?v=1">
<img
src="https://static.devin.ai/assets/gh-open-in-devin-review-light.svg?v=1"
alt="Open with Devin">
  </picture>
</a>
<!-- devin-review-badge-end -->

Author: spillai

tests/cli/test_cli_generate.py

@@ -7,24 +7,24 @@
 
 
 def test_generate_image(runner, mock_client, config_file, tmp_path):
-    """Test generate image command."""
+    """Test generate command with an image file."""
     path: Path = download_artifact(
         "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.invoice/invoice_1.jpg",
         format="file",
     )
     result = runner.invoke(
-        app, ["generate", "image", str(path), "--domain", "document.invoice"]
+        app, ["generate", "-i", str(path), "--domain", "document.invoice"]
     )
     assert result.exit_code == 0
 
 
 def test_generate_document(runner, mock_client, config_file, tmp_path):
-    """Test generate document command."""
+    """Test generate command with a document file."""
     path: Path = download_artifact(
         "https://storage.googleapis.com/vlm-data-public-prod/hub/examples/document.bank-statement/lending_bankstatement.pdf",
         format="file",
     )
     result = runner.invoke(
-        app, ["generate", "document", str(path), "--domain", "document.bank-statement"]
+        app, ["generate", "-i", str(path), "--domain", "document.bank-statement"]
     )
     assert result.exit_code == 0