Replace problem scale with solution space size

Brief notes:

1. Use n^k for basic variables.
2. Use (n+k-1)Unable to render embedded object: File (/(k-1)) not found. for chained and list variables.
3. Use logarithm to get the exponent instead of the exact number. It's faster.
4. Probably use logarithmic x-axis for the solution size in the benchmark report. Maybe switch between logarithmic and linear mode based on the difference between smallest and biggest data set.

Maybe use Stirling approximation for large factorials.