feat: harden CSV analysis parsing and schema inference

[rad1092]

Motivation

• Improve robustness of CSV ingestion and address incorrect dtype detection observed with pasted or semicolon/tab-delimited CSVs.
• Provide more accurate schema inference (distinguish int, float, date, string) so generated prompts contain better numeric context.
• Add standard deviation and unify file/text CSV parsing paths to eliminate inconsistencies between CLI and UI flows.

Description

• Added a dedicated _parse_csv_text that uses csv.Sniffer to auto-detect delimiters (,, ;, \t, |) with a safe fallback to , and used it for both text and file-based ingestion in bitnet_tools/analysis.py.
• Implemented stronger dtype helpers _to_int and _to_iso_date and expanded inference to emit int, float, date, or string per column, and compute std (population std) along with existing count/mean/min/q1/median/q3/max stats.
• Kept top-value ranking and missing counts logic, and ensured numeric stats are only emitted when all non-missing values are numeric.
• Updated tests (tests/test_analysis.py) to validate integer/date inference, presence of std, and semicolon-delimited CSV handling, and wired the CLI to use the improved parser (verified with an integration smoke run).

Testing

• Ran the unit test suite with python -m pytest -q which returned 3 passed.
• Performed a CLI smoke test with a semicolon-delimited CSV via python -m bitnet_tools.cli analyze /tmp/next.csv --question '요약' --out /tmp/next_payload.json and verified inferred dtypes were {'id': 'int', 'value': 'int', 'dt': 'date'}.
• Verified payload saving and prompt generation from both file and CSV-text code paths (payload JSON written and contains expected prompt template).

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fe775e6d32

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]

Reject non-finite floats during CSV numeric parsing

Parsing values with float(v) currently accepts tokens like NaN, inf, and -inf as numeric values, but those later cause statistics.pstdev to raise ValueError (e.g., "cannot convert NaN to integer ratio") when building numeric stats. This makes build_analysis_payload* fail for common CSVs that encode missing numeric values as NaN, breaking both CLI analyze and the /api/analyze UI flow instead of returning a summary.

Useful? React with 👍 / 👎.