fix(gateway): freeze-frame an entity that is serving last known values

[mfaferek93]

Closes #564.

content_has_live_data() rejected any payload whose entity reported
connected: false, so a loss-of-comms fault - the one fault whose freeze frame
is the whole point, showing what the device held right before it went away -
could never have one. The guard was aimed at the row of nulls a cold cache
yields, but that case is already caught by values_have_data() one step later.

The gate now judges the payload: a non-empty items array captures, whatever the
link flag says. captured_at_ns continues to state when the frame was taken, so
nothing is presented as fresher than it is, and a bridge that serves per-value
staleness keeps saying so on its own route.

Tests: DisconnectedEntityWithLastKnownValuesIsCaptured (new, fails before),
ContentHasLiveData.GatesOnItemsNotOnTheLinkFlag, and the two "nothing usable ->
no frame" tests re-pointed at the case they were really guarding (down and
empty). Full gateway suite: 3793 tests, 0 failures, 448 skipped.

[Copilot]

Pull request overview

Fixes entity freeze-frame capture for loss-of-communications scenarios in the gateway by allowing capture when an entity reports connected: false but still serves last-known values (i.e., non-empty items). This aligns freeze-frame behavior with the intended operational value of comms-loss faults: preserving the last observed state prior to disconnect, while still rejecting empty/all-null payloads.

Changes:

• Update EntityFreezeFrameCapture::content_has_live_data() to gate on a non-empty items array rather than the connected flag.
• Refine and expand unit tests to distinguish “disconnected with no usable values” (no frame) from “disconnected with last-known values” (frame captured).
• Update API/docs comments in the header to reflect the new gating semantics and the continued reliance on values_have_data() to reject all-null rows.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/ros2_medkit_gateway/test/test_entity_freeze_frame_capture.cpp	Adjusts existing tests and adds a new regression test to ensure disconnected entities with last-known values are captured.
src/ros2_medkit_gateway/src/entity_freeze_frame_capture.cpp	Changes content_has_live_data() to evaluate payload content (non-empty items) instead of the link state flag.
src/ros2_medkit_gateway/include/ros2_medkit_gateway/entity_freeze_frame_capture.hpp	Updates documentation of content_has_live_data() to reflect the new behavior and rationale.

[bburda]

Two things outside this diff that I think also need a change:

• Would you add an integration test for the loss-of-comms case? The route test plugin hardcodes {"connected", true}, so no launch test can reach the new path today. test_entity_freeze_frame.test.py already loads that plugin, so a second entity returning {"connected": false, "items": [...]} would cover it.
• Could you update the docs in this PR? docs/tutorials/snapshots.rst and docs/config/server.rst say the gateway snapshots the entity's "current data values" when the fault confirms, and docs/api/rest.rst calls freeze_frame "captured at fault confirmation". A frame may now hold values of unbounded age.

[mfaferek93]

Both outside-diff asks are in:

• test_route_data_plugin gained a second, external app serving {"connected": false, "items": [last known values], "timestamp": ...}; the new launch case asserts its fault carries those values plus the x-medkit connected / source_timestamp markers - 844f195.
• snapshots.rst, server.rst and rest.rst now say freeze-frame values may be last known when the entity is disconnected (age bounded only by the outage) and document the new x-medkit fields; captured_at is defined as dating the capture - daf5323.