computing-pi.md: example of JIT compilation not replicated

[wmotion]

I could not replicate the example proposed in https://esciencecenter-digital-skills.github.io/parallel-python-workbench/instructor/computing-pi.html#just-in-time-compilation-speedup

Calling the numbified function with an integer argument, then with a float, and again with a float produces comparable average elapsed times: 257 ns, 306 ns, 307 ns. (N = 10^6)

See screenshot below:

The explanation of the JIT compilation should be adjusted if this analysis is confirmed or its unexpected behaviour explained.

With numba 0.57.0, python 3.8.16

[wmotion]

This behaviour occurred also during the type-along in the edition of 4 July. As triaged then, this is strange also because sum_range_numba contains range(N) which should throw a fatal error when N is passed as a float. The computation proceeds instead, hence XPASS.

(We ought to check the function output to observe how Numba deals with this whole situation, that is, if it preserves correctness.)