Replace boost::random with std::random for WASM compatibility

- Remove boost::random dependency, use std::<random> and custom
  implementations for distributions not in the standard library
  (beta, laplace, logistic, pareto, rayleigh)
- Replace custom thread-local RNG with DuckDB's RandomEngine via
  RngAdapter (thread-safe, supports setseed() for reproducibility)
- Fix data race: acquire RandomEngine::lock before seeding
- Fix bug in DistributionCallUnaryUnary: args.data[2] -> args.data[1]
- Fix missing cauchy distribution registration
- Fix uniform_int to use boost::math::uniform_distribution<double>
- Fix quantile return types for discrete distributions
- Restore sampling for logistic, pareto, rayleigh distributions
- Remove stale forward declarations from utils.hpp
- Add BOOST_MATH_STANDALONE compile definition
- Add comprehensive test coverage for all 22 distributions
- Document setseed() reproducibility in docs/README.md
- Add .cache/ and vcpkg/ to .gitignore

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

Author: rustyconover

.gitignore

@@ -6,3 +6,5 @@ duckdb_unittest_tempdir/
 testext
 test/python/__pycache__/
 .Rhistory
+.cache/
+vcpkg/

CMakeLists.txt

@@ -21,7 +21,6 @@ file(GLOB DISTRIBUTION_SOURCES "src/distribution_*.cpp")
 
 set(EXTENSION_SOURCES
     src/stochastic_extension.cpp
-    src/rng_utils.cpp
     src/query_farm_telemetry.cpp
     ${DISTRIBUTION_SOURCES}
 )
@@ -31,6 +30,8 @@ build_loadable_extension(${TARGET_NAME} " " ${EXTENSION_SOURCES})
 
 target_link_libraries(${EXTENSION_NAME} Boost::math)
 target_link_libraries(${LOADABLE_EXTENSION_NAME} Boost::math)
+target_compile_definitions(${EXTENSION_NAME} PRIVATE BOOST_MATH_STANDALONE)
+target_compile_definitions(${LOADABLE_EXTENSION_NAME} PRIVATE BOOST_MATH_STANDALONE)
 
 install(
   TARGETS ${EXTENSION_NAME}

docs/README.md

@@ -224,6 +224,20 @@ Below are the parameters for each supported distribution. Use these as arguments
 | `min`     | Lower bound (integer) |
 | `max`     | Upper bound (integer, must be ≥ min) |
 
+## Reproducible Sampling
+
+All `dist_*_sample()` functions use DuckDB's built-in random engine. You can control the random seed for reproducible results using DuckDB's `setseed()` function:
+
+```sql
+-- Set seed for reproducible results
+SELECT setseed(0.42);
+
+-- These will produce the same values every time with the same seed
+SELECT dist_normal_sample(0.0, 1.0) FROM generate_series(1, 5);
+```
+
+Without calling `setseed()`, each session produces different random values.
+
 ## Usage Examples
 
 ### Normal Distribution