Implement a bunch of rhythm difficulty calculation fixes

[stanriders]

This PR addresses a couple of rhythm evaluation issues in osu!:

1. Removes island length size bonus completely and uncaps island size (was limited to 7 before)
2. Heavily nerf repeated islands such as triples in slowmotion (https://www.desmos.com/calculator/w9dtmvku2t)
3. Changes delta difference comparisons to use hitwindow-based epsilon
4. Buffs initial rhythm bonus
5. Fixes island comparisons to use delta as well as size
6. Considers doubletappness of islands
7. Refactor

Values can be checked here

[smoogipoo]

!diffcalc

[github-actions]

Target: #28871
Spreadsheet: https://docs.google.com/spreadsheets/d/1RjFSJwUQ_LW1wHNSeap2CwOXwwM9rm13hEqz_krZobY/edit

[smoogipoo]

!diffcalc

[github-actions]

Difficulty calculation failed: https://github.com/ppy/osu/actions/runs/10018306415

[apollo-dw]

LGTM

[stanriders]

@smoogipoo this change is good to go

[bdach]

@ppy/osu-pp-committee please advise how to fix merge conflict with #29980

...also #29998 touches this same thing again? what is going on here???

[stanriders]

@ppy/osu-pp-committee please advise how to fix merge conflict with #29980

Can use any value here, you can ignore any merge conflicts in Aim skillMultiplier as long as #29998 is the final pr that's getting merged. It was adjusted in half of prs to make individual changes more balanced in testing

[bdach]

?_?

i guess i'm going to merge #29998 into this instead if that is the case, after #29980 (comment) sheet is done

[smoogipoo]

Any reason for waiting on the sheet? I only requested a sheet to be queued, but is not required for merge.

[bdach]

well i mean if i run a sheet for the sake of running the sheet i can, but given that this touches the same code the preceding pull touched, it would make sense to verify that one doesn't do anything obviously broken before going forward? or not? but in that case we might as well run one sheet at the end of everything?

this is a complete mess to manage.

[smoogipoo]

If you're expecting branches to be working 100% as expected from the get-go, then sheets must look good before merge with no exceptions.

But, as we discussed, I'm fine not doing that in order to keep things moving at a rapid pace. My only requirement is for a sheet to be queued for each PR. It leaves a bit of a paper trail, if not to figure out where things went wrong / if a change has some edge case that was not known during rapid-merge, then to potentially be referenced if numbers are required to understand the change.

I will check every sheet myself, it's just for my own sanity :)

[bdach]

Alright well I'm merging master into this, then merging #29998 into this, then fixing tests, then queueing sheet. Means we can skip one sheet for #29998.

[bdach]

!diffcalc

[bdach]

Queueing again with explicit OSU_A spec since I realised this might not work as intended if I do it as above.

!diffcalc
OSU_A=cf66d40

[github-actions]

Target: #28871
Spreadsheet: https://docs.google.com/spreadsheets/d/1hzKVQpCxsOXM3T83nby6AhnLSifSlo0m8Y3hmIDT_zg/edit