I was mainly reacting about your remark:
"Though not sure about this because different coordinate systems on the same manifold can lead to different metrics!"
I was meaning that the information about the geometry that is contained in the metric tensor is independent of the coordinates used. (that's the meaning of the metric being a tensor)
Regarding the notion of parallelism, you need first to define what a line is, since parallelism, as far as I know, is related to "lines". In the context of Riemannian geometry (ie based on a metric), parallelism is defined on the basis of geodesics. The existence of parallel geodesics can be verified by exploring the metric tensor.
If the distinction between Euclidean, Hyperbolic of Elliptic geometry is defined by comparing the sum of the angles of a triangle to Pi, then the metric tensor is definitively able to detect this distinction. This is because, 1) the metric tensor is all that you need to define parallel transport and "lines" and 2) it defines the geometry in the tangent space (distances and angles).