feat: upgrade multi-CSV analysis with drift detection and group-ratio insights

[rad1092]

Motivation

• Move multi-file analysis beyond Excel-level summaries by adding distribution and comparison diagnostics so users can spot drift and subgroup composition quickly.
• Provide actionable follow-ups by generating Pandas example code and richer markdown reports to speed up exploration and visualization.
• Improve CLI ergonomics to enable grouped ratio analysis directly from multi-analyze so common workflows like 시도명 x 세차유형 are supported.

Description

• Enhanced bitnet_tools/multi_csv.py to compute numeric outlier ratio (IQR-based), dominant_value_ratio, and optional group-target ratio tables, and added cross-file _schema_drift summarizing dtype changes, missing-ratio range, dominant-value range, and numeric mean range.
• Extended bitnet_tools/analysis.py with a streaming summarize_reader to avoid full materialization and added build_markdown_report for single-file reports, plus robust numeric aggregation (count/mean/min/max).
• Updated CLI (bitnet_tools/cli.py) to add multi-analyze options --group-column and --target-column, added report, desktop, and doctor flows, and wire-ups to produce JSON + Markdown outputs via analyze_multiple_csv.
• Added small desktop UI (bitnet_tools/desktop.py + bitnet_desktop.pyw and BitNet_Desktop_Start.bat) and environment diagnostics (bitnet_tools/doctor.py), updated pyproject.toml to include bitnet-desktop, and expanded README examples to document the new grouped-ratio workflow.
• Tests: added/updated tests in tests/test_analysis.py and tests/test_cli.py to cover schema drift, group-target ratios, CLI behavior, and report generation.

Testing

• Ran pytest -q and all tests passed (13 passed).
• Verified CLI help shows multi-analyze and new options with python -m bitnet_tools.cli --help.
• Executed a sample multi-file run with python -m bitnet_tools.cli multi-analyze /tmp/g1.csv /tmp/g2.csv --question "확장분석" --group-column city --target-column type --out-json /tmp/g.json --out-report /tmp/g.md which produced the JSON and Markdown outputs successfully.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 36f4f3f81d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Escape newlines in generated pandas example

The pandas snippet emitted by build_code_guidance is syntactically invalid because \n is embedded directly inside single-quoted literals, so the generated text contains a real line break before the closing quote (e.g., in print('결측 비율 상위:\n', ...)). When users copy the pandas_example block from multi-analyze output and run it, Python raises SyntaxError: unterminated string literal, which breaks the new “actionable code guidance” flow.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Keep Tk widget reads on the main thread

_analyze_async starts a worker thread, but _analyze immediately reads Tk widgets (self.question.get, self.csv_text.get) from that background thread; Tkinter operations are required to run on the UI thread, so this can intermittently raise TclError or freeze/crash the desktop app when users click Analyze. Capture UI state on the main thread (e.g., via after) before dispatching background work.

Useful? React with 👍 / 👎.