[core] Mark allow-overlap symbols visible even outside of collision grid

[ChrisLoer]

Fixes issue #12683.

I tested the case this change was meant to address manually in a sample app that laid out a grid of allow-overlap: true symbols across the map. We don't have a good mechanism for automated testing of symbol fading on the native side, although when I do a port to GL JS I plan to write a test that will exercise the JS version of this logic.

cc @ansis @mkrussel77

[ansis]

This looks like this produces the right behaviour, but I'm wondering if it's possible to implement this without duplicating the text-optional logic.

Would making collisionIndex.placeFeature(....) always return placed: true for features with -allow-overlap: true achieve the same result? An extra condition to avoid adding offscreen features to the collisionindex might be necessary in this case.

[ChrisLoer]

Would making collisionIndex.placeFeature(....) always return placed: true for features with -allow-overlap: true achieve the same result? An extra condition to avoid adding offscreen features to the collisionindex might be necessary in this case.

I tried going down something like that path. The problem is that you have to apply the "optional" logic in parallel to both the "placed" result and the "outside the grid so don't insert into the collision index" result (whatever you call it). Although we're OK with having these allow-overlap symbols at the edge of the grid show without being inserted into the grid, in general the logic gets a lot harder to think about as soon as what's in the grid follows subtly different rules from what's showing. I may just not be thinking about it the right way, but my attempt ended up looking like significantly more complicated.

I was attracted to the per-bucket "always show" logic because at least the extra calculation only has to happen once per bucket.

[ansis]

I tried going down something like that path. The problem is that you have to apply the "optional" logic in parallel to both the "placed" result and the "outside the grid so don't insert into the collision index" result (whatever you call it). Although we're OK with having these allow-overlap symbols at the edge of the grid show without being inserted into the grid, in general the logic gets a lot harder to think about as soon as what's in the grid follows subtly different rules from what's showing. I may just not be thinking about it the right way, but my attempt ended up looking like significantly more complicated.

I was attracted to the per-bucket "always show" logic because at least the extra calculation only has to happen once per bucket.

Sounds good then!