Problem

When evaluating thousands of different condition sets against different facts using the same
default operators, users must create a new Engine() per condition set. This is expensive due to:

1. EventEmitter2 initialization per instance (Engine and Rule both extend EventEmitter)
2. 8 default Operator + 6 OperatorDecorator objects created per instance
3. OperatorMap with internal Maps allocated per instance
4. Rule/Condition tree parsed per addRule() call
5. Almanac with fact-caching Maps created per run() call

This results in ~20+ object allocations per evaluation. At scale (e.g. 250K evaluations),
this creates ~5M short-lived objects and significant GC pressure.

Solution

1. Shared default operator registry (singleton)

• Default operators and decorators arrays are now Object.freeze()d — created once at module load
• OperatorMap gains a static shared() method returning a singleton pre-loaded with all defaults
• OperatorMap supports a parent option for prototype-chain-style delegation
• Each Engine instance creates a lightweight child OperatorMap that inherits from the shared
  singleton instead of cloning all 16 operator/decorator objects

2. Static Engine.evaluate(conditions, facts, options?) method

A new lightweight evaluation path that bypasses all heavy infrastructure:

• No EventEmitter initialization
• No Rule construction or deep-cloning
• No event emission overhead
• Uses shared operator registry by default
• Supports full all/any/not nesting
• Returns API-compatible { results, failureResults, events, failureEvents }
• Accepts options: allowUndefinedFacts, pathResolver, operatorMap

3. Optimized Engine constructor

• Engine no longer creates 8 Operator + 6 OperatorDecorator objects per instance
• Creates an empty OperatorMap with parent: OperatorMap.shared(), inheriting all defaults
  via delegation — zero per-instance operator allocations for the common case
• Custom operators added to an Engine instance remain local and don't pollute the shared registry

Benchmark results (10K iterations)

new Engine() + addRule() + run() : 100.0ms (100,000 ops/sec)
Engine.evaluate() [static] : 34.4ms (290,908 ops/sec)
Reused engine.run() : 83.9ms (119,230 ops/sec)

Speedup (static vs new Engine) : ~2.9x

Changes