Add benchmark for rounding methods of FixedDecimal

[jedel1043]

This should help determine if the *_to_increment rounding methods affect performance or not.

Preliminary Results:

group	experimental	stable
rounding/ceil	1.00 704.3±13.32µs	1.03 725.9±16.01µs
rounding/expand	1.00 701.9±10.50µs	1.04 728.6±7.61µs
rounding/floor	1.00 712.1±21.57µs	1.02 723.6±14.02µs
rounding/half_ceil	1.02 789.2±13.48µs	1.00 773.7±9.05µs
rounding/half_even	1.01 769.4±19.95µs	1.00 764.8±3.59µs
rounding/half_expand	1.01 778.8±9.61µs	1.00 771.2±14.97µs
rounding/half_floor	1.01 782.9±19.59µs	1.00 772.9±13.17µs
rounding/half_trunc	1.00 771.0±8.55µs	1.01 778.4±14.95µs
rounding/trunc	1.00 696.2±13.94µs	1.02 707.0±10.35µs

[sffc]

Thank you! The benches look good so I will merge this and then run them locally

[sffc]

Here's what I get when enabling the experimental feature:

rounding/ceil           time:   [617.45 µs 686.12 µs 819.56 µs]
                        change: [+3.3587% +9.6679% +19.882%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 13 outliers among 100 measurements (13.00%)
  5 (5.00%) high mild
  8 (8.00%) high severe
rounding/floor          time:   [607.10 µs 609.13 µs 611.28 µs]
                        change: [-1.8892% +0.3084% +1.9135%] (p = 0.80 > 0.05)
                        No change in performance detected.
rounding/expand         time:   [585.07 µs 602.76 µs 629.84 µs]
                        change: [+0.6779% +2.6509% +5.5137%] (p = 0.01 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high severe
rounding/trunc          time:   [578.42 µs 602.10 µs 634.53 µs]
                        change: [-22.226% -11.365% -2.6846%] (p = 0.03 < 0.05)
                        Performance has improved.
Found 17 outliers among 100 measurements (17.00%)
  4 (4.00%) low mild
  1 (1.00%) high mild
  12 (12.00%) high severe
rounding/half_ceil      time:   [673.07 µs 678.14 µs 683.32 µs]
                        change: [-0.6328% +0.6679% +1.6434%] (p = 0.28 > 0.05)
                        No change in performance detected.
Found 21 outliers among 100 measurements (21.00%)
  5 (5.00%) high mild
  16 (16.00%) high severe
rounding/half_floor     time:   [668.83 µs 671.98 µs 675.82 µs]
                        change: [-2.6121% -0.2450% +1.2000%] (p = 0.85 > 0.05)
                        No change in performance detected.
Found 14 outliers among 100 measurements (14.00%)
  5 (5.00%) high mild
  9 (9.00%) high severe
rounding/half_expand    time:   [652.88 µs 655.40 µs 658.52 µs]
                        change: [+2.4361% +3.2498% +4.2415%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 5 outliers among 100 measurements (5.00%)
  3 (3.00%) high mild
  2 (2.00%) high severe
rounding/half_trunc     time:   [651.42 µs 654.50 µs 658.03 µs]
                        change: [-2.9493% +0.2538% +2.0996%] (p = 0.87 > 0.05)
                        No change in performance detected.
Found 16 outliers among 100 measurements (16.00%)
  4 (4.00%) high mild
  12 (12.00%) high severe
rounding/half_even      time:   [655.03 µs 671.18 µs 701.62 µs]
                        change: [-20.326% -8.5595% +1.1597%] (p = 0.20 > 0.05)
                        No change in performance detected.
Found 14 outliers among 100 measurements (14.00%)
  4 (4.00%) high mild
  10 (10.00%) high severe

Some a bit faster, some a bit slower, most in the noise.

I'm happy enough with these results to move forward with a single code path (with rounding increment) and deleting the old code path.