refactor tests

[danieleades]

small refactor of the tests.

None of these are integration tests. I've moved the performance tests to 'benches' and the remaining tests to unit tests within the lib.

It might be an idea to further refactor out the randomised test to a specific fuzzing target

[danieleades]

for discussion- i've removed the 'inline' attributes. This leads to slightly better performance on most bench tests on my machine:

Gnuplot not found, using plotters backend
shortest_csc_path       time:   [260.61 ns 261.85 ns 263.15 ns]                              
                        change: [-22.085% -17.785% -13.408%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  2 (2.00%) high mild
  7 (7.00%) high severe

shortest_ccc_path       time:   [187.12 ns 188.96 ns 191.87 ns]                              
                        change: [-2.6969% -1.0143% +0.6483%] (p = 0.25 > 0.05)
                        No change in performance detected.
Found 13 outliers among 100 measurements (13.00%)
  6 (6.00%) high mild
  7 (7.00%) high severe

shortest_path           time:   [354.46 ns 356.52 ns 358.73 ns]                          
                        change: [-15.823% -12.699% -9.7411%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  5 (5.00%) high mild
  4 (4.00%) high severe

many_sample             time:   [10.384 µs 10.846 µs 11.342 µs]                         
                        change: [+6.9659% +10.427% +13.981%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 7 outliers among 100 measurements (7.00%)
  6 (6.00%) high mild
  1 (1.00%) high severe

i suspect this is because the const methods can be more efficiently optimised if they're not inlined.

There is one performance regression here though- might be interesting in future to selectively replace the inline calls, verifying with the bench tests, and see which one is useful and which are counterproductive

[VirxEC]

Nice! I've been meaning to properly separate out the benchmarks.