feat: show banner when winml invoked with no arguments or --help (exit 0)

[hi-brenda]

Summary

When winml is invoked with no subcommand or with --help, the CLI should print a branded banner
to stderr and exit with code 0 (not 1).

Goal

Show a branded ASCII art banner when winml is invoked with no arguments or with --help.
Both cases are equivalent — they display the banner followed by Click's standard help output.

The no-args / --help terminal experience:

╭─────────────────────────────────────────────────────────╮
│                                                         │
│   __        ___       __  __ _       ____ _     ___     │
│   \ \      / (_)_ __ |  \/  | |     / ___| |   |_ _|    │
│    \ \ /\ / /| | '_ \| |\/| | |    | |   | |    | |     │
│     \ V  V / | | | | | |  | | |___ | |___| |___ | |     │
│      \_/\_/  |_|_| |_|_|  |_|_____  \____|_____|___|    │
│                                                         │
│   v0.0.1  ·  Windows ML model conversion and            │
│              optimization (CPU, GPU, NPU)               │
│                                                         │
╰─────────────────────────────────────────────────────────╯
Usage: winml [OPTIONS] COMMAND [ARGS]...

  WML ModelKit - Accelerate Model Deployment on WinML.

  Universal ONNX export with QNN and OpenVINO backend support.

Options:
  --version      Show the version and exit.
  -v, --verbose  Increase verbosity (-v=INFO, -vv=DEBUG)
  -q, --quiet    Quiet mode - errors only
  --help         Show this message and exit.

Commands:
  analyze       Analyze ONNX model for runtime support with live progress.
  build         Build a WinML-optimized ONNX model from a HuggingFace model
  compile       Compile ONNX model to EP-specific format.
  config        Generate WinMLBuildConfig for a HuggingFace model or .onnx f
  eval          Evaluate model accuracy on a dataset.
  expand_rules  Expand runtime rules zip files in-place when directories and
  export        Export HuggingFace model to ONNX format with HTP.
  hub           Browse ModelKit's curated built-in model catalog.
  inspect       Inspect input model's ModelKit configuration.
  optimize      Optimize ONNX model with capability-driven optimizer.
  perf          Benchmark model inference performance.
  quantize      Quantize ONNX model by inserting QDQ nodes.
  sys           Display system information for ModelKit export.

When it fires

• winml (no arguments) — banner printed to stderr, then Click's help output, exit 0
• winml --help — same behavior as above; both are equivalent

Never fires on any subcommand invocation (e.g. winml build --help).