Note: Also treats MultiThreshold as such a monotonic function, but this
needs more testing. Maybe MultiThreshold needs some special case
implementation more similar to the Quant operator?

Note: Does not cover the integer range analysis yet

Note: We should probably implement a specialized integer range
calculation for Split, but for now this seems to work just fine and
recovers the expected behavior from before the recent merge.

Note: Does not change the node execution behavior as this is already
covered by the assumptions of the multithreshold function, i.e.,
channels are axis 1.

Note: Only adds this to calc_intrange_eltwise_monotonic_intrangefirst
for now, which is only used for Split. The "scalebiasfirst" variant
could be extended in a similar way, but there is a warning in case of
multiple outputs - probably for some reason.