feat: write run_config.md alongside each analysis (#752)

[djconnexion77]

Closes #752.

Each saved analysis now includes a run_config.md at the report root, capturing:

• tradingagents package version (exposed via importlib.metadata)
• Ticker, analysis date, selected analysts, research depth, output language
• LLM provider, quick-thinking model, deep-thinking model, backend URL
• Max debate rounds, max risk discuss rounds, checkpoint flag
• Provider-specific reasoning params (OpenAI reasoning effort, Google thinking level, Anthropic effort) — only rendered when set

Why

Issue #752 describes comparing runs across LLM families (OpenAI vs Google vs Ollama) and over time as models and tradingagents itself evolve. Without persisted config, you can't tell from disk which parameters produced which report.

Implementation

• New tradingagents.__version__ via importlib.metadata.version("tradingagents")
• New cli/run_config_report.py with one pure function: build_run_config_markdown(selections, config, version) -> str
• cli/main.py:save_report_to_disk() gains optional selections/config kwargs; when both are passed, writes run_config.md. Backwards compatible — existing callers that don't pass them still work.
• 3 unit tests in tests/test_run_config_report.py (@pytest.mark.unit)
• CHANGELOG updated under [Unreleased] / Added

Sample output

# Run Configuration

| Field | Value |
|-------|-------|
| tradingagents version | 0.2.4 |
| Ticker | NVDA |
| Analysis date | 2026-05-04 |
| Analysts | market, news, fundamentals |
| Research depth | 1 |
| LLM provider | ollama |
| Quick-thinking model | qwen3:latest |
| Deep-thinking model | gpt-oss:latest |
| Backend URL | http://localhost:11434/v1 |
| Max debate rounds | 1 |
| Max risk discuss rounds | 1 |
| Checkpoint enabled | False |

Tests

$ python -m pytest -m unit
====================== 48 passed, 63 deselected in 1.24s =======================

(45 existing + 3 new for the markdown builder.)

[gemini-code-assist]

Code Review

This pull request introduces a run configuration report feature that captures the parameters used for each analysis, such as the package version, ticker, models, and provider-specific reasoning settings. The changes include a new run_config_report.py module, updates to the CLI saving logic in main.py, and the addition of unit tests. Feedback suggests moving local imports to the top of the file for PEP 8 compliance, using constants for the report filename to avoid hardcoding, and capitalizing labels in the markdown output for consistency.

[gemini-code-assist]

Local imports inside functions are generally discouraged unless they are necessary to avoid circular dependencies. Since tradingagents and cli.run_config_report are part of the project and don't seem to cause circularity here, moving these to the top of the file would improve readability and follow PEP 8 more closely.

[gemini-code-assist]

The filename run_config.md is hardcoded here and in save_report_to_disk. If the filename is ever changed in the saving logic, this print statement will become incorrect. Consider defining a constant for the filename or having save_report_to_disk return the names of all files it created to ensure consistency.

[gemini-code-assist]

For consistency with the other labels in the table (e.g., 'Ticker', 'Analysts', 'Research depth'), this label should be capitalized. Using 'TradingAgents version' would also match the project's branding in other parts of the CLI.

Suggested change

	_row("tradingagents version", version),
	_row("TradingAgents version", version),