M3-P1: Parallel workspace scan — Rust core with rayon

[hartsock]

Summary

Implement parallel workspace scanning using rayon. This is gitxtend's headline advantage over gila git-tend's sequential subprocess loop: N repos scanned in one process concurrently, with no fork-per-call overhead.

Blocked By

• M2-W1: Write-side Rust core — pull, push, add, commit #21 (M2-W1: pull/push/add/commit — needed because fetch is a write-side op used during scan)

Dependency Addition

[dependencies]
rayon = "1"

New File: src/scan.rs

use crate::status::{repo_status, RepoStatusData};
use rayon::prelude::*;
use std::path::PathBuf;

pub fn scan_repos(paths: &[PathBuf], fetch: bool) -> Vec<RepoStatusData> {
    paths
        .par_iter()
        .map(|p| repo_status(p, fetch))
        .collect()
}

Results are returned in input order (rayon preserves collect() ordering).

Modified Files

• src/lib.rs — add pub mod scan;
• Cargo.toml — add rayon dep

Tests

• src/scan.rs tests: scan 3 temp repos in parallel, verify all RepoStatusData returned
• Test mixed scenarios: some repos exist, some missing (error entries in result)
• Test fetch=false path (no network calls)

Acceptance Criteria

• cargo test passes
• Scan of 3 temp repos returns 3 RepoStatusData entries in input order
• cargo clippy -- -D warnings clean
• No PyO3 in src/scan.rs

Beaver (MacBook agent, Claude Sonnet 4.6)