fix: remove top-level concurrency block causing preflight deadlock

[jflowers]

Problem

The Release workflow's preflight job fails at startup with 0 steps executed
(null runner, inverted timestamps). This affects every release attempt —
both runs 32280364928 and 32288949495 failed identically.

Root Cause

The caller release.yml and the org-infra reusable preflight workflow
(reusable_release_preflight.yml) both declare:

concurrency:
  group: release-${{ github.repository }}
  cancel-in-progress: false

Both resolve to release-unbound-force/dewey. The caller run acquires
the group, then its child preflight job requests the same group —
creating a self-deadlock. GitHub resolves this by failing the reusable
job at startup.

Evidence

A/B comparison with replicator: replicator uses the identical org-infra
workflows at the same pinned SHA but its caller release.yml has no
top-level concurrency block — and its preflight ran all 9 steps
successfully.

Fix

Remove only the top-level concurrency: block (4 lines). The reusable
preflight already self-serializes via its own job-level concurrency group,
so the caller guard is redundant.

All other hardening is preserved:

• validate-branch job (main-branch enforcement)
• Dual-arch (amd64 + arm64) macOS signing and notarization
• Robust cask checksum patching with verification
• Strict semver (no allow_prerelease)

After Merge

Re-run: gh workflow run release.yml --repo unbound-force/dewey -f tag=v3.3.0

The preflight will detect v3.3.0 as a re-run (tag exists at HEAD), skip
creation, verify CI (build-and-test + MegaLinter are green on cd25cc0),
and proceed to GoReleaser.