Timestep calculation for CurvedMesh should be revised

I am no expert on this and whether it makes sense... @andrewwinters5000 would you mind having a look if this looks correct to you?

_Originally posted by @sloede in https://github.com/trixi-framework/Trixi.jl/pull/498#discussion_r606260489_

Start by reviewing the discussion linked from the issue, PR #498 discussion_r606260489, and inspect the CurvedMesh timestep calculation it refers to. The payload names no source file, test, or acceptance criteria, so confirm the intended calculation and expected result with the maintainers before making changes.

Written by the indexing model from the issue text.