feat(examples): add CSV data analysis agent example

[Mohammad Mohtashim (keenborder786)]

Adds a new data-analysis-agent example that analyzes CSV files using pandas
and matplotlib, producing structured reports with charts.

Use case as per in #951 to add more example

Why this example

The existing examples all use FilesystemBackend and return free-text responses.
This example fills two notable SDK gaps:

1. LocalShellBackend — first example to use shell execution via the
   execute tool. The agent writes and runs self-contained Python scripts
   (pandas, matplotlib) to compute statistics and generate charts, which is a
   fundamentally different interaction pattern from the file read/write workflows
   in other examples.

2. Structured output (ToolStrategy) — first example to use response_format.
   The agent returns a typed DataReport Pydantic model with findings, severity
   levels, and recommendations instead of unstructured text. This makes the output
   programmatically consumable.

It also uses memory (AGENTS.md) as enforceable analytical standards (rounding
rules, chart conventions) rather than persona context, and skills for three
distinct analysis workflows (exploratory, statistical, anomaly detection).

What's included

• analyze.py — async streaming with Rich progress display showing each tool call
• AGENTS.md — analysis standards (rounding, chart formatting, reporting rules)
• 3 skills: exploratory-analysis, statistical-summary, anomaly-detection
• Bundled data/sample_sales.csv (103 rows) for immediate demo
• README with installation, usage, and architecture docs

How I verified

• Ran the agent against the sample dataset with several queries
  ("Summarize this dataset", "Which region has the highest revenue?",
  "Find anomalies in daily revenue")
• Confirmed structured DataReport output renders correctly in the terminal
• Confirmed chart PNGs are generated in output/

Disclaimer: This contribution was developed with the assistance of AI agents.

[Mohammad Mohtashim (keenborder786)]

Hello Mason Daugherty (@mdrxy), I’ve added another example as requested in issue #951.

[Mason Daugherty (mdrxy)]

A couple things to align with the example contributing guidelines:

1. Commit uv.lock — the .gitignore currently excludes it, but all existing examples include their lock file for reproducible builds. Remove the uv.lock line from .gitignore and commit the generated lock file.

2. Tighten the deepagents version pin — deepagents>=0.3.5 has no upper bound, which could break on a future major release. Use a range instead, e.g. deepagents>=0.3.5,<0.4.0.