feat(pj_base): add DatasetDescriptor::requested_id for streaming lockstep

[pabloinigoblasco]

Summary

Add an optional requested_id field to DatasetDescriptor. When non-zero, the engine creates the dataset with that exact DatasetId (failing if it is already in use and advancing the counter past it); zero keeps the previous auto-assign behaviour, so existing callers are unaffected.

Why

This is the SDK half of a streaming secondary-engine id desync fix. The lockstep mirror needs to give the secondary engine the same DatasetId as the primary even when their id counters have drifted — e.g. a file loaded between two streams advances only the primary's counter, so a later createDataset on each engine would yield two different ids and the two engines would key the dataset differently.

The consuming side (DataEngine::createDataset honouring requested_id, plus the parallel TopicDescriptor::requested_id field and the streaming source manager wiring) lives in the app repo: PlotJuggler/PJ4#145. That PR does not compile against the current public SDK because the field added here doesn't exist yet — this PR unblocks it.

Scope

• One header change in pj_base/include/pj_base/dataset.hpp (+6 lines).
• Field is additive and default-zero: no existing callsite needs to change.
• No ABI break (struct grows by one DatasetId at the end; layout of existing fields unchanged).

Test plan

• CI green on all matrix jobs (header-only change; build only).
• PlotJuggler/PJ4#145 builds clean once this lands and the SDK Conan recipe is bumped.