Add default multiplier for mania key mods

[smoogipoo]

Resolves #30503

This doesn't go the full way. osu!stable will decrease the multiplier further if new_keys < old_keys, but this is hard to do here due to lack of ordering of calls to IApplicableToDifficulty.

• https://osu.ppy.sh/scores/3179253147

legacy: 859241
before: 959013
after: 863112

Those scoring values have been computed locally. Note that the "before" variant differs from the website because of a bug with not including mods correctly, that seems to have since been resolved but attributes haven't been reprocessed yet (#29772).

[bdach]

First of all:

!diffcalc
RULESET=mania
GENERATORS=score

Secondly: does this impact lazer scores at all? Because if it does, then a reverify is not enough. We'd need something like ppy/osu-queue-score-statistics#269.

[smoogipoo]

does this impact lazer scores at all?

Yes actually. That somehow totally slipped my mind.

[github-actions]

Target: #30506
Spreadsheet: https://docs.google.com/spreadsheets/d/1DdcdUi_BUcG3XSK92gNzMis92TyG7LaJBP7inGUnXto/edit

[peppy]

I'm have no strong opinions on this one, it looks to do what it says. @bdach any thoughts? Else I'd say we merge this in then figure out how to deal with it in reprocessing server-side as part of everything else score-related that needs reprocessing.

[bdach]

My unfiltered thoughts are that I'm thoroughly surprised that the series of changes that were supposed to facilitate mod multiplier recalcs (one of the pieces of which I've already linked above) has been sitting there half-reviewed for half a year and now we're apparently just gonna "do it live"? And if we're gonna do that then I'm going to go change classic mod multiplier to 1x outside of osu! right this second?

Like, there are considerations to be made on the multiplier recalc topic (e.g. ppy/osu-queue-score-statistics#269 (comment)) that I haven't heard any meaningful response to. So I'm not sure why it'd suddenly be fine to just "figure it out as we go".

[peppy]

I figured this was something we could deploy client-side and handle server-side later. If that isn't seen to be the case then it's probably best to hold off on some kind of coordination. This probably affects like 5 people anyway.

[bdach]

I obviously can't tell the scale of the issue, but I wouldn't say it's something that can be deployed to client not caring about server. It would stop the bleeding on it, but the messed up live leaderboards would remain messed up until corrected retroactively. Which, I'm not even currently sure we have the data to reliably do right now (is total score without mods stored to db for incoming scores? you'd also have to filter scores by build id to figure out which use the old and which the new multiplier). The open series of changes about mod multiplier recalculation is designed to remove such considerations.