Double check if there are no memory leaks or any performance hogs

[MrCsabaToth]

I wonder if this is some possible memory leak or similar bug. Things to check:

• Verify that surface creation is executed once and GPS array is populated only once, no duplicates: https://github.com/RecyclingTrashCans/recycling-trashcan-armap/blob/24ce063cd1c9f33b6ab69019a10cd5917d5c041c/app/src/main/java/dev/csaba/armap/recyclingtrashcans/TrashcanGeoRenderer.kt#L102
• Verify that we create the set of anchors and markers only once: https://github.com/RecyclingTrashCans/recycling-trashcan-armap/blob/24ce063cd1c9f33b6ab69019a10cd5917d5c041c/app/src/main/java/dev/csaba/armap/recyclingtrashcans/TrashcanGeoRenderer.kt#L227
• Verify that during rendering https://github.com/RecyclingTrashCans/recycling-trashcan-armap/blob/24ce063cd1c9f33b6ab69019a10cd5917d5c041c/app/src/main/java/dev/csaba/armap/recyclingtrashcans/TrashcanGeoRenderer.kt#L263 there's no memory leak
• Possibly there's no memory leak in the support / helper code https://github.com/RecyclingTrashCans/recycling-trashcan-armap/tree/main/app/src/main/java/dev/csaba/armap/common

[MrCsabaToth]

Ref https://github.com/google-ar/codelab-geospatial/issues/11 Demo videos https://youtu.be/w3dUH4fB68Y and https://youtu.be/pXJcbjEno_0

[MrCsabaToth]

The new version's 3D objects has way less vertex count (very rectangular arrow), so hopefully that'll help the 3D engine.

[MrCsabaToth]

The release version worked very smoothly on the Nord (although not in 100F heat, but screen recording was on still). So I repurpose this ticket.