I strongly suggest you write a function to cover your tiles with invisible objects, as tilemap collisions are really inefficient. That said, I might be persuaded to add a proof of concept "tilemap to invisible objects" script.
They are marginally faster (my way is really well-optimized), however they have a hard time allowing for slopes, moving (jump through) platforms and some of the other fancy stuff I'm doing. I reckon I could, but I am on the fence as I think the 10%+- performance gain that we can gain will complicate the engine by a lot. Additionally, as this runs at over 3000fps on my laptop, I think performance is not a problem right now.
It's not so much performance I was thinking of (but I know that tiles are faster than object collisions) it was more from a level / game design perspective where tiles are much easier to work with as an end user, especially with the autotiling stuff in GMS2 and ability to use things like Tiled and import maps directly into GMS2.
I'm not sure if your engine is a port from 1.4 code but it kinda feels like the old 1.4 way of doing things that hasn't been updated, maybe for compatibility reasons so the engine works with both 1.4 and 2.x which would be understandable.
Maybe the script(s) you mentioned to convert tiles / tile layers into the various invisible objects that your engine requires would be a good middle ground for everyone no matter whether they use 1.4 or 2.x?