Add M6 parallel-scaling coverage for 10+ concurrent subplans by
introducing a 15-subplan Behave scenario with explicit peak-concurrency
bound checks (max_parallel=10) and thread-safe concurrency tracking
via _build_executor().
Add deep hierarchical decomposition coverage (4+ levels) and adjust the
decomposition leaf condition to only stop early when hitting max_depth
or when the workset is trivially small (min_files_per_subplan). A
non-progress guard prevents pathological recursion when clustering
cannot meaningfully split the file set.
Add a small-project regression test (< 50 files) verifying that the
relaxed leaf condition does not cause unexpected decomposition depth
for small worksets.
Add ASV benchmark coverage for 15-subplan parallel execution to track
scaling behavior under higher concurrency.
Note: the _build_hierarchy child-linkage correctness fix (returning
node_id from recursive calls instead of using nodes[-1].node_id) has
been removed from this PR per review feedback — it is a separate bug
fix and will be submitted as an independent issue/PR per CONTRIBUTING.md
atomic commit policy.
ISSUES CLOSED: #855
brent.edwards