Microbench `BootOrchestrator::execute()` end-to-end (mocked)

[kat-perez]

Add a criterion microbenchmark covering the full BootOrchestrator::execute() path against mocked BootServices and RuntimeServices. Reports total cycles for the happy-path BDS sequence.

Why

The headline number for patina_boot perf. Lets us cite a single figure ("execute() takes N cycles") instead of diving into per-primitive detail. Comparing this number against the C BdsDxe equivalent is the simplest perf claim we can make to OEMs.

Acceptance criteria

• cargo bench --bench orchestrator runs and emits cycle counts for execute() (happy path, mocked deps)
• Output uses --output-format bencher so it's parseable by benchmark-action/github-action-benchmark
• Bench is wired into the PR-comment workflow once Workflow that runs benches + posts results as PR comment #122 lands

Dependencies

• Blocked by: Set up patina_boot/benches/ harness with criterion + mockall #115 (bench harness scaffolding must exist first)
• Blocks: Workflow that runs benches + posts results as PR comment #122 (PR-comment workflow has nothing to display until at least one bench exists)

[kat-perez]

The v22 execute_e2e benchmark is implemented and validated in draft PR #201, but it should not merge in its current raw-table form. Patina PR OpenDevicePartnership/patina#1743 removes StandardBootServices from the component model and introduces mockable UEFI service traits. #202 now tracks the required migration; after that lands, this benchmark should use generated service mocks. Runtime-variable access remains a separate boundary because #1743 does not yet provide equivalent runtime services.