Add granular I/O tracking for Windows with Python-level interception

[Copilot]

Windows lacked per-operation I/O data, preventing histogram visualization that Linux (strace) and macOS (fs_usage) provided.

Changes

Core Implementation

• Added _measure_windows_python_io() that intercepts builtins.open with an IOTracker wrapper class
• Wrapper tracks individual read(), write(), readline(), readlines(), writelines() calls and iteration
• Returns operation list with {'type': 'read'/'write', 'bytes': N} for histogram generation
• Updated _measure_linux_windows() to accept collect_ops parameter and delegate to tracking when enabled

Platform Coverage

• Linux: strace (system-level syscalls) or psutil fallback
• macOS: fs_usage (system-level file operations)
• Windows: Python-level tracking via open() interception

Limitations

• Captures Python open() calls only, not native C extensions
• Sufficient for pure Python workloads

Example

%%iops --histogram
with open('data.txt', 'w') as f:
    f.write('A' * 100)
    f.write('B' * 500)
    f.write('C' * 1000)

Now generates histograms on Windows showing operation count and bytes distribution by operation size.

Documentation

• Updated README with platform-specific tracking methods and Windows limitations

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Find a way to get granular (per operation) data on Windows</issue_title>
<issue_description></issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Find a way to get granular (per operation) data on Windows #7

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.