Refactor(rng): migrate core distributions from TRandom3 to std::mt19937_64

[HimanKing]

Summary

Partially migrates BioDynaMo's random number generation subsystem away from
ROOT's TRandom3 by replacing the backing engine with std::mt19937_64 and
re-implementing the majority of distributions using C++17 standard library
equivalents in src/core/util/random.h and src/core/util/random.cc.

Note: ROOT remains a required dependency. Landau, BreitWigner, and
all UserDefinedDist variants still delegate to ROOT internally. This PR is
a first, self-contained step toward full dependency reduction.

---

Motivation

ROOT is a large external dependency that significantly increases build
complexity, installation time, and reduces portability — particularly for
researchers without access to high-specification hardware or system
administration expertise. This refactoring targets the random number generation
subsystem as a self-contained, well-tested module suitable for incremental
dependency reduction.This is the first step toward breaking BioDynaMo's
dependency on ROOT, proving the migration path is viable one subsystem at a time.

---

Files Modified

• src/core/util/random.h
• src/core/util/random.cc
• test/unit/core/util/random_test.cc
• test/unit/core/simulation_test.cc

---

Changes

RNG Engine

• Replaced TRandom3 as the primary sampling engine with std::mt19937_64
  (mt_engine_).
• Added GetEngine() public accessor returning a reference to mt_engine_,
  allowing DistributionRng subclasses to draw samples without requiring
  friend access or exposing private members.
• SetSeed() seeds both mt_engine_ and the retained TRandom3 instance
  (generator_) with the same value, keeping unmigrated distributions
  reproducible under the same seed call.
• GetSeed() now returns last_seed_ — a uint64_t member set directly by
  SetSeed(). This removes the previous delegation to generator_->GetSeed()
  (TRandom3) and ensures the returned value is authoritative and independent of
  the ROOT engine, including if GetEngine() is used to interact with
  mt_engine_ directly.

Distributions Migrated to stdlib

Function	Original	Replacement
Uniform()	TRandom3::Uniform	std::uniform_real_distribution
Gaus()	TRandom3::Gaus	std::normal_distribution
Exp()	TRandom3::Exp	std::exponential_distribution
PoissonD()	TRandom3::PoissonD	std::poisson_distribution<long>
Poisson()	TRandom3::Poisson	std::poisson_distribution<int>
Binomial()	TRandom3::Binomial	std::binomial_distribution
Integer()	TRandom3::Integer	std::uniform_int_distribution
Circle()	TRandom3::Circle	trigonometric + std::uniform_real_distribution
Sphere()	TRandom3::Sphere	Marsaglia method (std::normal_distribution)

Distributions Not Yet Migrated (future work)

• Landau() — no standard library equivalent; still delegates to
  generator_->Landau(). Requires a custom rejection-sampling implementation.
• BreitWigner() — no standard library equivalent; still delegates to
  generator_->BreitWigner().
• UserDefinedDistRng1D/2D/3D — user-defined distributions backed by
  ROOT's TF1/TF2/TF3, sampled via GetRandom(). These use ROOT's global
  gRandom engine internally and are unaffected by SetSeed() on a Random
  instance. A full replacement requires a rejection-sampling framework and is
  deferred.

Architectural Note — Stale Virtual Interface

The DistributionRng<T>::SampleImpl(TRandom* rng) virtual interface retains
its ROOT-typed signature. Implementations that have been migrated to stdlib
silently ignore the rng parameter (marked /*rng*/ in source). This
interface will be updated to remove the TRandom* parameter once all
distributions are ported.

---

Testing

All unit tests pass under the refactored implementation alongside the full
suite of system-wide unit tests: