revert: fix: throw when LatLng is non-finite

[JaffaKetchup]

Reverts #2182 (which is fdc089a)

Fixes #2199. Reopens #2178.

[mootw]

looks good. maybe we look at resolving the re-opened issue on the marker layer level.

Are there any places internally in FM that generate NaN's, or is this purely an end user inputs bad data causes weird behavior downstream?

[mootw]

it looks like final newOffset = offset * (1.0 - 1.0 / scale); is a likely culprit. maybe it is a damper function attempt but it is undefined at zero, which causes NaNs. I wonder if we rework that code to avoid Nan generation

[mootw]

offset is just a scalar (not included here)