fix(project): trim whitespace from project names

[baiirun]

Problem

Project names were stored exactly as given. A padded value such as -p " spindle" (for example from a shell variable that zsh didn't word-split) created a separate project that looks identical in prog show but never matches prog list -p spindle, because every filter compares names exactly. In practice, one epic and its tasks were split across " spindle" and spindle.

Change

• model.NormalizeProject trims surrounding whitespace. It's applied on every write path: the CLI -p flag, prog project, CreateItem, SetProject, and EnsureProject.
• Schema migration v4 trims existing names in items, learnings, labels, and concepts, and merges padded projects rows into the trimmed name. It runs once when prog next opens a database. Every statement can be re-run safely. A label or concept whose trimmed name already exists is left as is, so the migration doesn't fail on it.

Why a migration instead of whitespace-insensitive filtering

Loose matching would still leave duplicate projects, labels, and learnings under two spellings. Normalizing once at the storage layer keeps exact matching correct everywhere.

Testing

• TestCreateItemNormalizesProjectForFiltering: a task created with " brandnew\t" is found by -p brandnew.
• TestMigrateTrimsExistingProjectNames: seeds a mix of padded and clean rows at v3, migrates twice, and checks filtering and that no padded projects remain.
• Both tests fail without the fix. go test ./... and go vet ./... pass.
• Checked against a copy of a real database: list -p spindle went from 5 items to all 13, with a single spindle project and user_version 4.

🤖 Generated with Claude Code