MediaApp: expand benchmarking to cover full pipeline timing

[N0153128]

Benchmarking currently tracks single-file processing time only. Need visibility into the full pipeline broken down by phase.

Requested changes

Three timing phases to track:

1. Core processing — startup time covering OpenCV load, plugin registration, CLI parsing (tracked in Main)
2. Per-file overhead — time spent in MediaApp before processor.process() is called: MIME detection, plugin resolution, context population, validation
3. Plugin processing — time spent inside processor.process() only

Reporting:

• Per-file breakdown logged after each file when benchmarking is enabled
• Running average updated after each file in a batch
• Final average printed after all files are processed

Implementation notes

• Core startup time: capture Instant before plugins.registerAll() and after cli.execute() returns in Main
• Per-file overhead vs plugin time: capture Instant before and after processor.process() call inside processFile() — delta between total file time and plugin time gives overhead
• Average tracking: keep a running total and file count in MediaApp, expose a summary method called after the loop in DisarmCLI

Acceptance criteria

• Core startup timing captured in Main
• Per-file overhead and plugin processing time tracked separately in MediaApp
• Running average maintained across multiple files
• Summary printed after batch completes
• All output gated behind globalConfig.getBenchmarking()