Add dedicated screenshot subcommand for page captures

[cschanhniem]

Feature Request

The current CDP escape hatch (browser-control cdp Page.captureScreenshot) works but requires the user to know CDP internals and manually handle Base64 decoding, format selection, and file I/O.

A dedicated screenshot subcommand would make this common operation a first-class citizen.

Proposed API

# Full-page screenshot, auto-detect format from extension
browser-control screenshot page.png

# Set format and quality explicitly
browser-control screenshot --format jpeg --quality 80 page.jpg

# Capture a specific region (x, y, width, height)
browser-control screenshot --clip 100,50,800,600 region.png

# Write Base64 to stdout (for piping back to agent context)
browser-control screenshot --stdout > screenshot.b64

Rationale

Coding agents frequently need to capture page state for visual context — for example after a click or fill to confirm the result. A dedicated command would:

1. Handle Base64 decoding and file writing internally
2. Support --format png|jpeg|webp with quality control
3. Support --clip for region captures (useful for isolating UI elements)
4. Support --fullpage via Page.captureFullScreenshot for tall pages
5. Optionally write Base64 to stdout for agent tool-use workflows without temp files