[SDE] Port CaoTauLeaping implementation from StochasticDiffEq.jl#603

[ChrisRackauckas]

Migrated from PR: SciML/StochasticDiffEq.jl#603
Author: @sivasathyaseeelan

A complete CaoTauLeaping implementation was contributed but not merged before the SDE migration. This PR contained:

• Proper adaptive step size selection for CaoTauLeaping using Cao et al. (2006) pre-leap τ computation
• Separate CaoTauLeapingCache / CaoTauLeapingConstantCache (previously shared TauLeaping's cache)
• Enhanced TauLeaping with adaptive flag and post-leap error estimates
• Full docstrings with references
• Tests in test/tau_leaping.jl

Files changed: src/algorithms.jl, src/caches/tau_caches.jl, src/integrators/stepsize_controllers.jl, src/perform_step/tau_leaping.jl, test/tau_leaping.jl

Full diff preserved at: https://github.com/SciML/StochasticDiffEq.jl/pull/603.diff

Reference:
Cao, Yang, Gillespie, Daniel T, Petzold, Linda R. "Efficient step size selection for the tau-leaping simulation method", J. Chem. Phys. 124(4), 044109 (2006).