AP_DDS: support automatic reconnect to micro-ROS agent

[srmainwaring]

Add ping test and attempt to automatically reconnect to the micro-ROS agent.

Closes #23372

For details see: #23372 (comment)

[Ryanf55]

So, it works great on multiple interface cycles. Super fantastic improvement to the behavior.

[Ryanf55]

Can you use PONG_IN_SESSION_STATUS instead of magic 1, which is set in session.h of the Micro-XRCE-DDS-Client library>

[srmainwaring]

38f5310

[Ryanf55]

This is set from uxr_init_session(&session, comm, key);, so perhaps better wording would be session invalid?

[srmainwaring]

Think the wording is ok: comm is set when the transport initialises - before the session is initialised.

bool AP_DDS_Client::ddsSerialInit()
{
    // setup a framed transport for serial
    uxr_set_custom_transport_callbacks(&serial.transport, true,
                                       serial_transport_open,
                                       serial_transport_close,
                                       serial_transport_write,
                                       serial_transport_read);


    if (!uxr_init_custom_transport(&serial.transport, (void*)this)) {
        return false;
    }
    comm = &serial.transport.comm;

[Ryanf55]

Do you intend to set a timeout of 0 here? Note - for a non-obvious C-api like this one, I might actually just create some constexpr locals to show what is being passed in.

constexpr int timout_ping_ms {0};
constexpr int ping_attempts {1};
uxr_ping_agent_session(&session, timout_ping_ms, ping_attempts);

[srmainwaring]

Added const stack variables in:
a14e0af

Didn't add constexpr as we're adding for readability and are happy for the compiler to optimise these away.

The 0 timeout looks odd doesn't it? This is consistent with the approach used by the PX4 DDS implementation - it must be trying to poll at least once otherwise I'd expect a timeout. Need to trace this more thoroughly to understand why it does not fail every time.