I agree, this seems to be more mathematical problem than programming one. There was also no mention of optimalization part of challange. I did simplest iteration based fibonacci calculation and got timeout when n was beyond 1 mil. I don't understand why there are tests with n bigger than 1 mil, the name suggest to get proper 1 mil value. I'm kinda dissappointed with this kata as I did not want to specifically search for mathematical ways of getting different equation to solve this without using the series generation.

Why would it have to? It's somewhat clear from the definition of Fibonacci series, and is mentioned in HINT I given in the description.
It's not that the author decided it to be this way, it's a consequence of applying the formula and extending it into negative indexes in a way explained in the description, isn't it?