Use `stacklevel` parameter when emitting warnings.
Warnings do not print a backtrace, only a single line of code. By default that's the line of code where the warning is emitted, typically somewhere deep inside the Kwant source.
warnings.warn
supports a stacklevel
parameter that allows to display some point that is higher in the backtrace instead. We should use this parameter consistently such that users are pointed to the line in their code that triggered the warning. Currently this is not the case for all the warnings in Kwant.