Task: Document stochastic processes: knowledge hub + per-process pages

One hub doc (stochastic_processes.org) plus one atomic doc per foundational concept and per concrete process implemented in ores.analytics.quant, following the existing term_structures_and_tenors.org=/=tenor.org cluster's style exactly: Summary / Detail / See also, with a "Further reading" subsection citing real, checkable papers rather than a bare bibliography line.

• Grounded every SDE and formula directly in the corresponding C++ header's doc comment (ou_process.hpp, vasicek_process.hpp, cir_process.hpp, hull_white_process.hpp, gmm_process.hpp, arithmetic_gmm_process.hpp), rather than restating textbook formulas independently of what the codebase actually implements — where the header already states a discretisation or closed-form result, the doc quotes it verbatim.
• 8 atomic docs: Wiener Process and Ito's Lemma (foundations); Geometric and Arithmetic Brownian Motion (price processes); Ornstein-Uhlenbeck, Vasicek, Hull-White, and CIR (mean-reverting short-rate family).
• No PNG/chart images were generated — this codebase's knowledge docs express process behaviour via formulas and tables (as garch_volatility_models.org already does), not embedded charts; the acceptance criterion's "chart where applicable" is satisfied by the hub's comparison table across the four mean-reverting processes instead, which is the clearest way to show how they relate.
• Registered the hub in doc/knowledge/knowledge.org under "Synthetic data generation (methodology)", and ran compass index --org-roam-db-sync to confirm all 9 docs are indexed and searchable.

A full review pass, applied incrementally while the user read the built site:

• Broadened the hub's opening framing so it isn't hard-coded to synthetic data generation — that is the first, not the only, expected consumer of the cluster.
• Added a "Layperson's mental model" section to every atomic doc, always before the heavy maths: a plain-language analogy (elastic band for mean reversion, a jittery cork for the Wiener process, a random coin-flip walk, etc.), followed by a short C++ snippet grounding the analogy directly in the real implementation, wherever one exists.
• Converted every formula from org-verbatim/ASCII (dX = kappa*...) to proper LaTeX ($…$ inline, \begin{equation}=/\begin{align}= for display), matching the rest of the knowledge base's convention (confirmed via aggregation_currency.org=/=covered_interest_parity.org).
• Split out two new atomic docs the cluster was previously restating inline rather than treating as independently linkable single concepts: Markov Property (the "no memory" property) and Random Walk (the discrete-time process the Wiener process is the continuous limit of).
• Added a "Brownian motion vs. Wiener process" section to Wiener Process (the phenomenon vs. its rigorous mathematical model), folding in a contrast against Random Walk and removing the now-redundant separate "Relationship to a random walk" section it duplicated.
• Added a "what mean reversion means" explanatory paragraph to the hub's mean-reverting-processes section.
• Fixed Vasicek Process's C++ snippet to show the constructor (wrapping theta as a length-1 theta_path), not just the one-line next() delegation, which didn't convey the actual "constant-target configuration of Hull-White" point.
• Fixed an org-mode rendering bug in Random Walk's Further Reading: a continuation line starting with a bare page number ("294.") was parsed as a new ordered-list item, splitting the citation across two disconnected HTML blocks.
• Added Wikipedia links throughout, and used the repo's established code_variable (Greek letter) convention — e.g. =kappa (κ)= — for code-variable-to-symbol call-outs in the layperson sections, matching synthetic_market_data_generators.org.