Add distributional arithmetic prototype

[valayDave]

Motivation

• Provide a small, self-contained prototype that models scalar random variables as lazy expression graphs so dependence (shared vs independent sources) is explicit.
• Offer a minimal API for building random-variable expressions, evaluating them with reproducible particle sampling, and inspecting summary statistics and event probabilities.
• Create a platform to iterate toward planned features such as symbolic simplifications, FFT convolution and planner-based evaluation strategies.

Description

• Add a new distarith package (distarith/__init__.py) implementing Distribution subclasses (NormalDistribution, LogNormalDistribution, EmpiricalDistribution) and a RandomVariable expression DAG with Constant, Source, UnaryExpr, and BinaryExpr nodes.
• Implement arithmetic and comparison operators on RandomVariable to produce lazy expression graphs and support iid()/independent_copy() to create independent copies of a source while preserving source identity by default.
• Provide particle-based evaluation with RandomVariable.evaluate(...) / sample(...), summary helpers mean(), variance(), std(), quantile(), an EvaluationResult container, and Event/CompoundEvent semantics with the helper P(...) for probability estimation using joint samples.
• Add convenience constructors Normal(...), LogNormal(...), Empirical(...), and unary transforms exp/log, plus type hints and basic input validation.

Testing

• Ran python -m pytest test/unit/test_distarith.py, which executed 4 unit tests and all tests passed.
• Ran python -m py_compile distarith/__init__.py test/unit/test_distarith.py to verify byte-compilation, which succeeded.
• Unit tests exercise source-identity semantics (x - x), independent copies (iid()), joint-event probability evaluation (P(profit < 0)), empirical quantiles, and compound event probabilities, and all assertions passed.

---

Codex Task