Use magic for binary detection

[andrew]

Use github.com/git-pkgs/magic to classify files before packing. Preserve the existing binary skip result and accept only UTF-8 text, while catching binary signatures, disallowed controls, and NUL bytes beyond the old 8 KiB probe.

Add coverage for UTF-16, malformed UTF-8, late NULs, binary formats without NUL, Unicode, and empty content.

[Copilot]

Pull request overview

This PR updates Outline’s packing pipeline to use github.com/git-pkgs/magic for content classification so that only UTF-8 text is packed while reliably skipping binary content (including files with binary signatures, disallowed control bytes, malformed UTF-8, and late NULs beyond the prior 8 KiB probe).

Changes:

• Replace the previous NUL/UTF-8 prefix probe with magic.Detect-based text gating via isPackableText.
• Expand packing tests to cover PDF detection, UTF-16, malformed UTF-8, disallowed controls, late NUL bytes, Unicode, and empty content.
• Add the github.com/git-pkgs/magic v0.1.0 module dependency.

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated no comments.

File	Description
pack.go	Switches file packability checks from a fixed-size binary sniff to magic-based classification while preserving the "binary" skip reason.
pack_test.go	Adds/updates tests to validate the new classification behavior across more encodings and binary/text edge cases.
go.mod	Introduces the github.com/git-pkgs/magic dependency.
go.sum	Adds checksums for the new magic dependency.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.