perf(reference_picker): implement incremental parsing

[sudo-tee]

This PR is to address some performance issues in the rendering of large sessions

[Copilot]

Pull request overview

This PR targets performance in the reference picker flow by switching reference_picker.parse_references() to an incremental, per-message cache and updating unit/replay fixtures accordingly. It also updates the reference struct fields (e.g., col instead of column) and adjusts picker collection logic to source references from the parse cache.

Changes:

• Implement incremental parsing + caching for code references keyed by message_id.
• Update reference struct fields and navigation behavior (including missing-file warnings).
• Update/replace unit tests and large replay expected JSON fixtures to match new output.

Reviewed changes

Copilot reviewed 13 out of 14 changed files in this pull request and generated 5 comments.

File	Description
lua/opencode/ui/reference_picker.lua	Reworks parsing into an incremental cached scanner; updates picker collection + navigation.
tests/unit/reference_picker_spec.lua	Updates assertions and adds tests for caching/incremental behavior + picker collection from cache.
tests/data/*.expected.json	Refreshes replay fixture outputs (mostly formatting/serialization changes).
.claude/settings.local.json	Adds local Claude permissions configuration.

---

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