Reduce map rebuild work during live tracking

[robekl]

Problem

The live map path rebuilt from broad app-state notifications, including frequent timer-driven updates unrelated to map data. It also recreated network tile providers during rebuilds, causing avoidable work during active tracking.

Change

Narrowed the map's subscriptions to map-relevant state only, introduced a dedicated revision signal for map data mutations, and reused tile providers instead of recreating them on each rebuild.

Why This Fixes It

The map now rebuilds when map inputs actually change rather than on every unrelated provider notification. Reusing tile providers also avoids repeated client allocation. Together, those changes reduce runtime churn on the hottest UI path during wardriving.

[MrAlders0n]

Hey! Great optimization, context.select() is the right call here. A few things worth looking at if you have time:

1. The 17-field anonymous record type is copy-pasted in three places, so if one changes they all need to change. Mind pulling it into a typedef MapState = (...) at the top of the file?
2. _baseTileProvider and _overlayTileProvider get created in initState() but never disposed. Pretty sure CancellableNetworkTileProvider holds HTTP clients so those should probably get .dispose() calls in dispose().
3. The repeater MarkerLayer block looks like it picked up a couple extra spaces of indentation compared to the ones above it.
4. A doc comment on _markMapDataChanged() would be nice, just so future contributors know that forgetting to call it means the map silently shows stale data.

Also, are you on our Discord?

[robekl]

That all makes sense. PR updated.

[446564]

please resolve conflicts, likely of note is that dev is now formatted with dart format ., so ensure your changes are formatted correctly to pass CI.