feat: add zero-config agent regression workflow

[Therealdk8890]

Summary

Adds a shorter first-use workflow for agent regression testing:

python agent.py
dpk record

# After making a change
python agent.py
dpk compare
dpk gate

Adds dpk as a short alias for dprovenancekit.
Adds zero-argument record, compare, and gate commands.
Allows traced_run(context_id="...") to automatically persist runs locally.
Atomically pins known-good runs to .dprovenance/baseline.sqlite.
Preserves provenance edges and existing explicit database/run-ID CI workflows.
Updates documentation, changelog, and regression coverage.

Why
The existing regression engine was capable but required users to configure stores, database paths, and run IDs before seeing value. This change reduces the path from installation to a meaningful agent regression check while retaining the advanced configuration options.

Validation
492 tests passed, 3 skipped
Query parity suite passed
Ruff, mypy, and compileall passed
Standard and adversarial corpora remain at 1.000 precision, recall, and F1
Verified from a clean Python 3.13 wheel installation
Confirmed dpk gate exits with status 1 when the candidate drops its verification step