FENCE-2804: Port bubble-geofence + region-removal methods to Swift

[stevepopovich]

Summary

Continues the RadarLocationManager ObjC→Swift migration by porting the bubble-geofence + region-removal cluster, following the established beacon-sync seam pattern.

Ported as @objc static twins on RadarLocationManagerSwift:

• replaceBubbleGeofence(locationManager:location:radius:)
• removeBubbleGeofence(locationManager:)
• removeAllRegions(locationManager:)

Each ObjC method dispatches to its twin when useSwiftLocationManager is enabled and returns; both implementations coexist until the Swift port is trusted. Instance state (CLLocationManager) is passed in explicitly.

Changes

• RadarLocationManager+Swift.swift — three Swift twins + mirrored radar_ / radar_bubble_ prefix constants.
• RadarLocationManagerSwift.h — ObjC-visible declarations.
• RadarLocationManager.m — useSwiftLocationManager dispatch shims (ObjC bodies kept for coexistence).
• project.pbxproj — registers the new test file.
• RadarLocationManagerSwiftRegionTests.swift — behavior tests: prefix-scoped removal, tracking-gated add, and the new region's center/radius.
• AGENTS.md — note that make format reformats the entire repo (separate commit).

Manual test steps

Part A — Example app setup

1. Force the Swift path on. Hardcode RadarSdkConfiguration.swift:60 so server config can't flip it mid-test:

   useSwiftLocationManager = true   // was: dict?["useSwiftLocationManager"] as? Bool ?? false

2. Build & run on a real device (bubble geofences need real Core Location). Grant "Always" location. In the app's Logs tab, ensure debug logs are visible (Radar.setLogLevel(.debug) if needed).

Part B — Verify the Swift path (flag ON)

1. Tap startTracking with the Responsive preset (it enables both the stopped and moving bubble geofences at a 100 m radius).
2. On the first location update, confirm the emoji-prefixed line appears — this proves the Swift twin ran:
   • 🦅 Successfully added bubble geofence | latitude = …; longitude = …; radius = 100; identifier = radar_bubble_<uuid> (from replaceBubbleGeofence).
   • The absence of the 🦅 on this line would mean the ObjC body ran instead → fail.
3. Move / re-evaluate: walk ≳150 m (beyond the 100 m radius) so the SDK syncs and replaces the bubble. Confirm:
   • 🦅 Removed bubble geofences (from removeBubbleGeofence), immediately followed by a new
   • 🦅 Successfully added bubble geofence | … at the updated location.

Part C — Regression check (flag OFF, ObjC path)

1. Flip RadarSdkConfiguration.swift:60 to useSwiftLocationManager = false, rebuild, and repeat Part B.
2. Behavior should be identical, except the same log lines now appear without the 🦅 prefix (Successfully added bubble geofence | …, Removed bubble geofences) — confirming the ObjC fallback does the same work.

Cleanup

Revert the RadarSdkConfiguration.swift hardcode before committing.

[nickdonnelly]

Maybe we make these @objc static lets and ref them from the objective-c codebase so we don't have to keep them in sync?

[nickdonnelly]

or reference them in swift via a shared header that exposes them

[stevepopovich]

I think this is ok. Sharing them makes it a bit more work to clean it up and I would like to keep the implementations as separate as possible. Usually I would agree with you but in this transitional case, I think it is ok to have some duplication so this file is fully testable without using obj c code

[nickdonnelly]

I'd prefer filtering before the loop instead of using continue to skip iterations.

let validUUIDs = uuids.prefix(numUUIDs).compactMap { uuid -> (String, UUID)? in
    guard let proximityUUID = UUID(uuidString: uuid) else {
	   // log here
	   return nil
   }
   (uuid, proximityUUID)
}

// Then this, or use a `.forEach` since you know you won't need to bail on any of the iterations
for (uuid, proximityUUID) in validUUIDs {
	// work
}

[stevepopovich]

Tomato, tomato 🙃

I feel like having having 2 loops is marginally slower and unneeded

[nickdonnelly]

I'd have more sympathy for the performance argument if this wasn't bounded to a 9 elements 😅
But you're right, it's purely a personal preference nit - functional composition over collections reads easier to me at a glance.

[stevepopovich]

I'd have more sympathy for the performance argument if this wasn't bounded to a 9 elements

Good point 😆