Make step text opt-in

[kanishkpaul]

What this changes

Makes typed and pasted content opt-in in the one-shot replay path. It also aligns get_steps with get_activity defaults, lets callers reuse a custom min_minutes selection, and orders replay events by timestamp.

Why

The README promises output-level content opt-in, but get_steps previously returned captured text by default. It also rebuilt a different default time window from get_activity, which made a frame id harder to reproduce.

User impact

• MCP and CLI step expansion now return character counts by default.
• Pass include_text: true or --include-text to request capped typed/pasted content.
• get_steps and aframes steps now default to a two-hour lookback and accept min_minutes / --min-minutes for a matching selection.
• The legacy --no-text flag remains accepted.

Deferred

This is intentionally a compatibility-focused first fix. Exact multi-monitor event provenance, opaque stable frame references, and guarded workflow execution need a separate schema/design change.

Validation

• pytest -q (108 passed)
• python -m build --sdist --wheel
• Added regressions for MCP and CLI text opt-in, legacy CLI opt-out, selection parameters, and out-of-order event insertion.

Checklist

• Output stays deterministic (no clock, network, randomness, or model calls)
• New behavior has a test; pytest -q passes
• No typed text or page content is emitted by default