feat: implement Phase 1 — ternary probabilistic modelling core

Rewrite manifest to parse betlangiser.toml with probabilistic variable
declarations (Normal, Uniform, Beta, Bernoulli, Custom distributions),
simulation config (samples, confidence, seed, output format), and full
parameter validation.

Implement ABI types: Distribution with validated constructors,
TernaryBool with complete Kleene strong three-valued logic (AND/OR/NOT/
implies), ConfidenceInterval, SimulationConfig, and SimulationResult.

Add codegen submodules (parser, distribution, codegen) that generate
Betlang source code with distribution declarations, uncertainty
propagation rules, ternary logic helpers, and Monte Carlo entry points.

84 tests pass: unit tests for all ABI types, codegen modules, manifest
validation, and integration tests covering init, generate, end-to-end
pipeline, De Morgan's laws for ternary logic, and all distribution types.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: hyperpolymath

Cargo.lock

@@ -0,0 +1,39 @@
+# betlangiser manifest — Risk model example
+# SPDX-License-Identifier: PMPL-1.0-or-later
+#
+# This example demonstrates ternary probabilistic modelling with betlangiser.
+# Three variables model a simplified pricing scenario:
+# - price: normally distributed around 100 with std_dev 5
+# - demand: uniformly distributed between 50 and 150 units
+# - will-buy: a Bernoulli variable representing purchase likelihood (70%)
+#
+# The generated Betlang code wraps these in probability distributions,
+# applies Kleene ternary logic to the Bernoulli variable, and runs
+# a Monte Carlo simulation to estimate confidence intervals.
+
+[project]
+name = "risk-model"
+description = "Probabilistic pricing model with ternary purchase logic"
+
+[[variables]]
+name = "price"
+distribution = "normal"
+mean = 100.0
+std-dev = 5.0
+
+[[variables]]
+name = "demand"
+distribution = "uniform"
+min = 50.0
+max = 150.0
+
+[[variables]]
+name = "will-buy"
+distribution = "bernoulli"
+probability = 0.7
+
+[simulation]
+samples = 10000
+confidence = 0.95
+seed = 42
+output-format = "text"