Safer and more efficient GitHub content fetching #92

Summary
Fetching a single file from GitHub clones/pulls the whole repository branch, which is slow and brittle. Improve safety, speed, and resilience.

Proposed changes

• Prefer lightweight fetch for single files:
  • Use raw.githubusercontent.com or the GitHub Contents API for HTTPS where possible.
  • Support pinning by tag or commit SHA in addition to branch.
• Add network robustness:
  • Timeouts, retries with exponential backoff.
  • Configurable retry/timeout via CLI/env.
• Security controls:
  • Optional allowlist/denylist of protocols.
  • --deny-network flag to disable all remote fetching.

Files

• struct_module/content_fetcher.py
• docs/file-handling.md (document protocol behavior and options)

Acceptance criteria

• Single-file fetch no longer requires a full repo clone by default.
• Pinning by commit or tag is supported and documented.
• Timeouts/retries configurable and covered by tests.
• Backwards compatibility maintained for existing configs.