[dombrovsky]

Extends the already-merged 1.0.0 benchmark baseline from #50 with four keyed-scheduling cases and three bounded-admission cases, for 23 modern cases total. The benchmark project remains separate from standard solution tests and package publication; production and existing unit-test sources are unchanged.

The history/comparison runner now uses an explicit allowlist of the nine common harness files. Newer-API sources are excluded from both release and candidate checkouts, preserving the 16-case release-compatible contract and the existing middleware exclusion for older releases. The standalone suite and Benchmarks workflow still cover all 23 modern cases.

The default baseline is the exact 1.0.0 release commit 0800af84229d7beb3a6fa507d236d0ff533b0c92, with production-tree verification and the common harness overlaid in an independent checkout. This removes reliance on the deleted benchmarks-1.0.0 branch and its unreachable benchmark commit after rebasing.

The README documents the existing history and same-run comparison workflows, durable records on benchmark-data, and current reproduction commands. Raw measurements remain outside source control.

Validation:

• All 12 Python reporting/regression tests pass, including modern/future-source exclusion, missing common-source rejection, and release CLI baseline selection.
• Solution restore, whitespace formatting and verification, and file-scoped namespace verification pass.
• End-to-end runs against exact 1.0.0 and current main built successfully and validated all 16 common cases twice on each version.
• Sequential burst measurements on the same runner completed on both versions with three warmups and eight actual samples; report parsing and identical common-harness hashes were verified. These runs validate the measurement pipeline, not a performance regression verdict.
• The unchanged modern benchmark sources previously passed all 23 workload validations and GitHub benchmark smoke checks; fresh CI will rerun on this update.

Closes #43 (parent #42).