v1.7.1-embark

SCICO-1241: address codex review feedback

Three independent issues caught by a second-pass review of the fix:

1. Drop libboost-dev from the Dockerfile. boost::dynamic_bitset is the
   only boost dependency and a vendored copy ships under include/boost,
   wired in via the Makefile's -I include. The apt install was dead
   weight; the previous build success on the GHA runner was masked by
   ubuntu-24.04 having libboost-dev preinstalled.

2. Make Individuals::closeOutputFileHandles exception-safe and call it
   explicitly from GERMLINE::mine. Previously it ran inside ~Individuals
   only; if sortFileInPlace threw (e.g. sort(1) exited non-zero on disk
   full or path quoting), the destructor propagated and the program
   terminated mid-teardown with map entries still pointing at deleted
   ofstreams. Now the destructor wraps closeOutputFileHandles in a
   try/swallow safety net, the explicit drive-by from mine() runs on
   the clean path so failures surface as runtime_error, and
   closeOutputFileHandles itself stores the first sort exception,
   completes every close + clear, and rethrows at the end.

3. Correct the Match.cpp comment about the new-vs-new branch. It is
   not dead code — Share::assertMatches falls back to "all pairs"
   when hasRestrictions() is false (i.e. when -samples_to_compare_to
   is unset). The oline.clear() there is a real fix for a latent NF=14
   bug, not just defense in depth.

No behavior change on the success path. Existing tests still pass
byte-exact (host + linux/amd64 ubuntu:24.04 builder). Verified the
docker image rebuilds cleanly without libboost-dev and that
make test stays green at 1.7.1-embark on the dev server.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

v1.7.0-embark

(no-ticket): don't auto-mark releases as latest

softprops/action-gh-release defaults to make_latest: true when the
input is unset, and we were also setting it explicitly. Flip to
make_latest: false so the "Latest" tag on the releases page has to be
updated manually — this branch's -embark fork releases out of band
from upstream and we don't want CI racing a release into the
"Latest" slot.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

v1.6.0-embark

(no-ticket): tag current branch