In #32973, #32981, and #32995 some improvements to the --warn-long feature are discussed. Particularly in #32981, the second_on_modern_computer() multiplier is dropped, because as it stands, it often disables the --warn-long warnings entirely.

To reinstate the "CPU handicap," we should replace that function with one that performs a fixed amount of work. The amount of CPU time taken to perform that work can then be used as a multiplier to tweak the default --warn-long values.

This idea is still not perfect, but it should be better than a constant value, and avoids the problems with using old doctest statistics.

CC: @tornaria

Component: doctest framework

Issue created by migration from https://trac.sagemath.org/ticket/33022