[Alexander-Mitrofanov]

This PR preserves 173 historical validation and experimental source files that remained outside the tracked application worktrees. They are stored under doc/local-validation-archive/files/ with their original relative paths and SHA-256/Git blob hashes. A README describes their scope and provenance, and the source distribution includes a verified sources.zip containing the same 173 files. This preserves long historical paths under the existing tar format.

The code comprises standalone header/installed-library consumers, dependency/configure probes, benchmark and native-output comparison harnesses, plotting/report scripts, instrumentation patches, and alternate source controls used during earlier measurements. Examples include a matrix prototype, dense/hash partition experiments, an uncached kinetic control, and historical compiler-compatibility copies. The copied sources are inactive archival material: they are not application replacements or build/test/CI targets.

Why it was not committed before: these helpers and controls lived in issue-specific validation/profiling directories rather than the tracked source checkouts. Earlier PRs committed the selected production changes, while exploratory variants and local verification tooling stayed outside Git. Not every file has a separate decision record, so this PR does not claim every experiment was an overlooked fix. It commits that material now for Martin's inspection before the local workspace is removed.

The C++23 assessment and deliberately set-aside compiler proposal have a separate PR (#277). The actual six-file ensemble-memory integration port has a separate production PR; this archive excludes that port's files. The misleading deletions in issue-260-validation/debug-master came from a copied .git pointer against a later index: they are not proposed as production deletions. Its older header is retained only as a historical source snapshot.

Byte-identical content already reachable from published Git refs is omitted and recorded in the manifest. Generated builds, dependency downloads, private correspondence, delivery payloads and large raw evidence are outside this source PR; a verified local recovery archive outside the IntaRNA folder preserves remaining workspace material.

Validation: all 89 archived Python/shell files pass syntax checks; every stored file matches its recorded source bytes and hash; make dist succeeds and every one of the 173 sources extracted from the distributed ZIP matches its recorded SHA-256; staged whitespace checks pass. Old C++ experiments were not rebuilt against current master, and no new scientific or performance claim is made. This is intentionally a draft archival PR so preserving the record is separate from choosing any experiment for production.