* feat(uxrce_dds_client): accept RTCM corrections from ROS 2
ROS 2 setups that need RTK had to keep a MAVLink link beside DDS just for GPS_RTCM_DATA. Routing by device_id gives each DDS source its own rtcm_corrections instance, so two publishers never interleave their byte streams. The DDS side needs no framing: the injectors reframe the stream from arbitrary chunks.
Assisted-by: Claude:claude-opus-5-5
* docs(releases): note RTCM corrections over uXRCE-DDS
Assisted-by: Claude:claude-opus-5-5
* fix(uavcan): bound RTCM corrections length before forwarding
uXRCE-DDS is the first rtcm_corrections source that does not bound len. A ROS 2 RtcmData with len over 300 made the DroneCAN forwarder read past the data array and broadcast that memory as RTCMStream.
Assisted-by: Claude:claude-opus-5-5
* docs(msg): drop the uXRCE-DDS routing note from RtcmData
It described the client's routing rather than the message, and the device_id field comment already covers it.
Assisted-by: Claude:claude-opus-5-5
* fix(uavcan): bound moving-baseline length before forwarding
The forwarder trusted the publisher's len, the same gap just closed for rtcm_corrections. No current publisher exceeds the array, so this only keeps a future one from reading past it.
Assisted-by: Claude:claude-opus-5-5
---------
Co-authored-by: Beniamino Pozzan <beniamino.pozzan@gmail.com>