feat(dc_mcap_writer,docs): add MCAP as a passthrough Destination

[Minipada]

Summary

• Records can now be recorded as .mcap files for ros2 bag info/Foxglove replay. Vector has no MCAP sink, so per ADR-0003 this needed a blessed-vs-passthrough decision recorded before implementing — docs/adr/0009-mcap-as-a-passthrough-destination.md records the choice: passthrough, not a dc_bridge-side blessed Destination. The issue's stated justification for the Bridge-side option (local-first durability surviving a Shipper outage) isn't a real gap: the Shipper is already at-least-once and disk-buffered per blessed sink (ADR-0002).
• New package dc_mcap_writer (ament_python, no ROS deps of its own): a standalone process that consumes the public dc.<tag> routes over a Vector socket sink (NDJSON) and writes rotated .mcap files, one JSON-schema Channel per Tag (the foxglove/mcap jsonschema/writer.cpp pattern the issue links). Rotates by size or duration, whichever comes first.
• Demo (dc_demos/config/mcap_sink.toml, params/mcap_recording.yaml, launch/mcap_recording.launch.py) and a new tutorial (doc/src/dc/demos/mcap_recording.md) following the existing Elasticsearch/InfluxDB passthrough demo shape.
• mcap added as a Python dependency (pyproject.toml/requirements.txt), matching how numpy/opencv-python/pydantic/Pillow are already managed outside rosdep.

Test plan

• 12-case pytest suite for dc_mcap_writer (rotation by size/duration, per-Tag channel routing, all three time_format conversions, malformed-NDJSON handling, mcap round-trip via mcap.reader) — all pass in a throwaway venv (mcap/pytest installed via uv, no ROS 2 available in this environment)
• CLI --stdin smoke test end-to-end: piped NDJSON in, read the resulting .mcap back independently, confirmed Channels/schema/timestamps round-trip correctly
• pre-commit run against every changed/new file (black/isort/pyupgrade/flake8/xmllint/yaml/toml clean; pycln/poetry-requirements fail in this environment the same pre-existing way earlier work already logged in progress.txt)
• colcon build --packages-select dc_mcap_writer and ros2 bag info/Foxglove against a real recorded file — needs a ROS 2 Jazzy environment (the podman-based dc-workspace image used for prior C++ verification), not available in this session

Closes #210

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 40.81%. Comparing base (588c495) to head (e7c394f).

Additional details and impacted files

@@           Coverage Diff           @@
##            jazzy     #319   +/-   ##
=======================================
  Coverage   40.81%   40.81%           
=======================================
  Files          82       82           
  Lines        4919     4919           
=======================================
  Hits         2007     2007           
  Misses       2912     2912           

Flag	Coverage Δ
cpp-jazzy	40.81% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.