Fix unit test failures: pygments lazy module access and PyTorch 2.6 compatibility

[Copilot]

Why are the changes needed?

Two categories of unit tests were failing:

1. test_source_code_renderer - lazy module attribute access doesn't support nested paths
2. PyTorch tests - PyTorch 2.6 changed torch.load default from weights_only=False to True

What changes were proposed in this pull request?

1. SourceCodeRenderer pygments import

File: plugins/flytekit-deck-standard/flytekitplugins/deck/renderer.py

Replace lazy module attribute chain with direct imports:

# Before
formatter = pygments.formatters.html.HtmlFormatter(style="colorful")

# After  
from pygments import highlight
from pygments.formatters.html import HtmlFormatter
from pygments.lexers.python import PythonLexer

formatter = HtmlFormatter(style="colorful")

Aligns with pattern already used in flytekit/deck/renderer.py.

2. PyTorch backward compatibility

Files: flytekit/extras/pytorch/native.py, flytekit/extras/pytorch/checkpoint.py

Added explicit weights_only=False to torch.load calls:

return torch.load(local_path, map_location=map_location, weights_only=False)

Maintains compatibility with models/checkpoints saved before PyTorch 2.6. Includes security warnings about arbitrary code execution risk when loading untrusted pickle files.

How was this patch tested?

• plugins/flytekit-deck-standard/tests/test_renderer.py: All 10 tests pass
• tests/flytekit/unit/extras/pytorch/: All 13 tests pass
• CodeQL security scan: 0 alerts

Setup process

N/A

Screenshots

N/A

Check all the applicable boxes

• I updated the documentation accordingly.
• All new and existing tests passed.
• All commits are signed-off.

Related PRs

N/A

Docs link

N/A

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.