feat: implement cdd scan

[hammadmajid]

What

Adds internal/scan.Run(ctx, cfg config.Config, hist *history.History) (Summary, error), per the resolution in #13 and #9.

• Discovers Projects under cfg.Root with cfg.Exclude and cfg.IncludeHidden via project.Discover.
• For each Project, computes an evidence time: git.LastCommit when it is a repository with a commit, else the Project directory's mtime (truncated to UTC second precision to match History's line format, so repeated Scans compare equal).
• Calls hist.Seed(project.Rel(), evidence) for each Project. Seed's own idempotency rule (never overwriting a newer real Visit) is not re-implemented here.
• Bounds concurrent git calls to a worker pool of 8.
• Returns Summary{Seeded, Projects int} so the cli can print "Seeded 12 Visits across 40 Projects".

Why

Closes #19. A fresh cdd installation has no History until a Jump happens; Scan seeds one Visit per discovered Project so the Picker has a sensible starting order.

Tests

Table of scenarios in internal/scan/scan_test.go, using t.TempDir() Roots and real git repositories:

• a repository Project is seeded from its commit time
• a non-repository Project is seeded from its directory mtime
• an already-jumped Project is left alone
• a second Run seeds nothing new
• an excluded Project is skipped

Notes for reviewer

• Summary.Seeded is derived by comparing hist.Count before and after each Seed call, rather than re-implementing Seed's idempotency rule in this package.
• Git calls are gathered concurrently (bounded to 8 workers) before any Seed calls, which run sequentially; Seed itself is already safe for concurrent use (it takes an advisory flock), so this ordering is purely about keeping Summary.Seeded accounting simple, not about correctness.

Closes #19

🤖 Generated with Claude Code