Scope repoclosure's persistent-repo exclude to the package's own binary names

[ogajduse]

Summary

--newest --best (#436) resolves Requires against the newest package across all repos
combined, while still checking every repo's own newest package in isolation. A subpackage that
self-pins via Requires: %{name} = %{version}-%{release} (e.g. a -doc/-cockpit sibling)
then false-positives the instant a newer build of the same SRPM lands anywhere else in the
checked set — the persistent staging repo's stale copy has nothing left to satisfy its exact
pin against.

Reproduced against the real theforeman/foreman-packaging#13921 and #13922 CI failures
(rubygem-foreman_remote_execution and rubygem-foreman_leapp bumps), using the live
el9-foreman-plugins-nightly-staging repo and the real repoclosure/yum.conf.

Fix

Exclude the package under test from the persistent staging repos it's checked against, since
the PR's own freshly-built repo (repo0/downloaded_rpms) is authoritative for it. Scoped
with dnf's per-repo --setopt=<repo>.excludepkgs=, never applied to the fresh repo itself.

Exclude list is resolved to the package's exact binary names via rpmspec, never a
name-prefix glob — a glob would also match unrelated packages sharing a name prefix (e.g.
foreman* also matching foreman-proxy, rubygem-hammer_cli* also matching
rubygem-hammer_cli_foreman), silently masking a real cross-package regression on those
packages — exactly the failure mode --best (#436) exists to catch
(theforeman/foreman-packaging#13647/#13661/#13688). Falls back to the exact hostname
(never a glob) if the spec fails to parse; resolves to no exclusion at all for repoclosure-only
hosts with no package directory.

Testing

• New unit tests for repoclosure.py's extracted build_command().
• New/updated functional tests covering: single and multiple persistent target repos, the
  downloaded_rpms release flow, repoclosure-only (spec-less) hosts, and a package with an
  intentionally unparseable spec (exercises the exact-hostname fallback branch).
• Full suite: 135 passed. pylint: 10.00/10.
• Manually re-validated against real dnf4 (matching the CI environment) with synthetic and
  live-repo reproductions of both the original false positive and the cross-package regression
  --best must keep catching.

🤖 Generated with Claude Code

[Odilhao]

Approved, tested locally and it worked against the PRs that are broken atm on foreman-packaging