perf(rules): optimize ruleset scope partitioning

[tryhard-26]

during ruleset initialization in capa.rules, _get_rules_for_scope was called separately for all 8 scopes. inside that loop, it iterated over every non-subscope rule and invoked get_rules_and_dependencies, which rebuilt index_rules_by_namespace and the rule dictionary from scratch on every call. across the default 1,054 rule files (1,397 rules in memory), this triggered 8,440 redundant namespace index builds and repeated full dependency traversals, followed by 8 separate topologically_order_rules sorts on the same rule set. this created an o(scopes * rules^2) bottleneck that added roughly 3.5 seconds of overhead to every ruleset instantiation.

this change adds _get_rules_by_scope to traverse the dependency graph and topologically sort reachable rules in a single pass across all scopes, then slices each scope from that pre-ordered list. get_rules_and_dependencies and topologically_order_rules now accept optional pre-indexed namespace and name mappings to avoid redundant dictionary allocations.

benchmarked across the default 1,054 rule files (1,397 rules in memory) in an identical harness:

metric	before	after	speedup
scope rule partitioning	3,540.5 ms	7.5 ms	472.1x
total ruleset init	3,580.8 ms	46.7 ms	76.7x

a new regression test verifies that per-scope rule membership and topological dependency order invariants hold across the full ruleset.

• No CHANGELOG update needed

[github-actions]

Please add bug fixes, new features, breaking changes and anything else you think is worthwhile mentioning to the master (unreleased) section of CHANGELOG.md. If no CHANGELOG update is needed add the following to the PR description: [x] No CHANGELOG update needed

CHANGELOG updated or no update needed, thanks! 😄