Made significant optimisation changes across all features

[Attakay78]

perf: eliminate O(N) hot-path bottlenecks for large route counts

Fixes 8 scaling bottlenecks that degrade linearly with route count. No public API changes.

Changes

File	Change	Before → After
engine.py + models.py	Cache GlobalMaintenanceConfig in-process; use frozenset for exempt-path checks	1 backend read + O(M) list scan per request → cached; O(1) lookup
middleware.py	Pre-build static/param route lookup cache at startup	O(N) route walk per request → O(1) dict lookup for static paths
file.py	Write-through in-memory cache; immediate flush for state, debounced (50ms) for audit	Full file read+write per operation → zero reads after warm-up
redis.py	Replace KEYS scan with SMEMBERS on a dedicated route-index set; per-path audit lists	Blocking KEYS scan → safe SMEMBERS; fetch-1000-filter-in-Python → fetch exactly limit
memory.py	deque + per-path audit index	O(N) audit scan per filtered query → O(k)
openapi.py	Cache filtered schema keyed to engine._schema_version	O(N×verbs) rebuild per /docs request → rebuilt only on state change
engine.py + router.py	register_batch(): one list_states() read for all routes at startup	N get_state() reads at startup → 1 list_states() read