Replace error functions in DifficultyCalculationUtils with good-enough approximations

[Natelytle]

For an acceptable loss in precision, we can speed up the error function implementations by around 5x. This affects both standard and taiko, as both modes use DifficultyCalculationUtils.Erf for their accuracy scaling. In standard, this necessitates a change to the maximum tested precision, however taiko's calculation is not precision-sensitive and is therefore fine without any changes.

This massively speeds up the standard performance calculator, by around 2x.

Old: New:

Erf approximation delta visualization: https://www.desmos.com/calculator/lyg8opgpej

ErfInv approximation delta visualization: https://www.desmos.com/calculator/npbyvxpwtw

[tsunyoku]

!diffcalc RULESET=osu OSU_A=https://github.com/ppy/osu/tree/pp-dev OSU_B=https://github.com/ppy/osu/pull/33717 SCORE_PROCESSOR_A=https://github.com/tsunyoku/osu-queue-score-statistics/tree/add-hd-so-to-relevant-legacy-mods SCORE_PROCESSOR_B=https://github.com/tsunyoku/osu-queue-score-statistics/tree/add-hd-so-to-relevant-legacy-mods

[tsunyoku]

!diffcalc RULESET=taiko OSU_A=https://github.com/ppy/osu/tree/pp-dev OSU_B=https://github.com/ppy/osu/pull/33717

[github-actions[bot]]

Target: https://github.com/ppy/osu/pull/33717 Spreadsheet: https://docs.google.com/spreadsheets/d/1VvdZ4z5X424U0o7rFtvEd_fQqZ_pzDis9NOzrhJnv3M/edit

[github-actions[bot]]

Target: https://github.com/ppy/osu/pull/33717 Spreadsheet: https://docs.google.com/spreadsheets/d/1Ujsxvmw7YIxo1ZiZg6Rtcx3mE6Hi1Nk3UHyeFCICNuo/edit