chore(release): @hasna/repos v0.1.42

[andrei-hasna]

Ships the foreign-clone resolution fix from #65 as @hasna/repos 0.1.42.

Release span — the whole span, not just one PR

d823955 (the 0.1.41 release commit) → f9ebad5:

efa0e63  fix(repos): never resolve a remote to a _factory_src scratch clone (#65)
f9ebad5  chore(release): @hasna/repos v0.1.42

One behaviour-bearing commit. No other surface is touched.

What ships

repos repo --remote hasnaxyz/iapp-takumi --json returned the _factory_src mirror at rc=0 with checkout_health: usable, on a tree dated 2026-05-24 — the deterministic lookup every global agent rule points at, handing back a two-and-a-half-month-stale clone as success.

getRepoByRemote now drops foreign copies — a separate clone with its own object store and its own HEAD — before every other rule. A worktree is deliberately not foreign: it is another view of the same clone, so a live worktree still wins over a hollow primary.

Full reasoning, measurements and the two-sided regressions are on #65.

Gates on the merged tree

CI on #65                Typecheck, Test & Build   pass 1m9s
local bun run typecheck  rc=0, no diagnostics
local bun test           791 pass / 1 pre-existing docs-parity timeout,
                         which fails identically on the unmodified base tree
secrets scan             0 hits, control fires on a known-positive

Refs: todos c0ac7e9b-8df1-4cd6-9e65-55e3b626b8f0 (A3-00208)

Agent: vespasian

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}