[datlechin]

Why

When a plugin is both bundled and installed in the user folder, discovery picked the copy with the higher CFBundleShortVersionString. A stale registry copy of a bundled driver could beat the built-in one; five bundled plugins carry hand-set marketing versions to prevent it.

Change

• Discovery picks the copy with the higher CFBundleVersion, read from disk. A tie goes to the built-in copy, and the losing user copy is pruned as before.
• PluginEntry carries the build.
• Deleted the unused bundleVersion(at:) and the one-time com.TablePro.needsRestart cleanup.

Verified

New PluginWinnerSelectionTests (lower build loses, higher wins, tie goes to built-in) and bundleBuild(at:) cases in PluginBundleVersionTests.

Not in this PR

• Every plugin build is 1 until bundles are stamped with the app's build, so until then the built-in copy wins every tie, including over a newer hand-installed zip of a bundled driver.
• Docs: development/testing-plugins.mdx:21 still says the higher version wins at discovery. It becomes "the higher build (CFBundleVersion) wins" in the docs PR that rewrites that page for the v3 manifest.
• PluginManager.init's userDefaults parameter is now unused. Dropping it edits PluginManager.shared, the line next to refactor(plugins): stamp the PluginKit version from one build setting #3498's change, and PluginLazyActivationVerificationTests, so it follows refactor(plugins): stamp the PluginKit version from one build setting #3498.