feat: Add LLMSimulationTraceCompressor for token-efficient AI summaries

[dyrpsf]

Overview

This PR introduces the LLMSimulationTraceCompressor to the org.simulator.math package. This is a foundational utility for the upcoming gsoc-sysbio-llm-tools architecture, designed to specifically solve the LLM token-limit bottleneck when handling massive simulation outputs.

Changes Made

• Data Compression: Built a utility to intercept MultiTable time-series data and algorithmically extract critical biological waypoints (initial states, peak concentrations, peak times, and final steady states).
• Token Optimization: Formats the extracted data into a highly dense, LLM-readable summary string, preventing context-window overflow and reducing AI hallucinations.
• Testing: Added LLMSimulationTraceCompressorTest.java utilizing Mockito to mock MultiTable behaviors and verify the statistical extraction logic.

Validation

• Compiles cleanly within the sbscl module.
• Validated locally alongside the full SBSCL test suite.

[Schmoho]

I would likely not frame this specifically around LLMs but more generally as a summary string mechanism.

As for the test, I am uncertain whether there is enough covered with that. At least there is a test though, thank you!

[dyrpsf]

I would likely not frame this specifically around LLMs but more generally as a summary string mechanism.

As for the test, I am uncertain whether there is enough covered with that. At least there is a test though, thank you!

You make a great point about keeping the core library naming generic. I have renamed it to SimulationTraceCompressor so it isn't strictly tied to LLM use cases. I also added some extra assertions to the test suite to cover edge cases like empty traces!

[draeger]

Minor note; I like to advice using 1d, 1f, 1L, etc. to indicate the type of numeric value when working with literals.

[draeger]

I think it is more precise to write 1d or 1f, etc. rather than 1.0. The latter will certainly be interpreted as a double but the first two versions give us developers freedom to specify how much memory we want to spend.