Improve fullcombo threshold formula

[Givikap120]

It used to always take 10%.
Now it's down to 5% if map has no hard sliders, since in this case probability of dropping many sliderends is low.
For maps with hard sliders the portion can reach up to 20%.

[stanriders]

!diffcalc
RULESET=osu
OSU_A=https://github.com/ppy/osu/tree/pp-dev
OSU_B=#35724

[tsunyoku]

!diffcalc
RULESET=osu
OSU_A=https://github.com/ppy/osu/tree/pp-dev
OSU_B=#35724

[github-actions]

Target: #35724
Spreadsheet: https://docs.google.com/spreadsheets/d/1vmwziNTzMhI0jnrtpVyf1prgIR2ZdxhshangfUm8uz0/edit

[Givikap120]

Those are results (green are correct prediction, red is incorrect):

I think it should be made harsher for high top weighted slider factor values.
All the red scores on second screenshot are stream maps with very big amount of sliderends missed.

[tsunyoku]

!diffcalc
RULESET=osu
OSU_A=https://github.com/ppy/osu/tree/pp-dev
OSU_B=#35724

[github-actions]

Target: #35724
Spreadsheet: https://docs.google.com/spreadsheets/d/1NTayPISUkS0FH8cbXvY_qaCqO5QPbujU2cPy7QY2orc/edit

[Givikap120]

Updated.
First three green greens scores are sliderbreak on the slider right before the final spinner, unlucky, missing only 3 combo.
Two new "red" maps in the first screenshot are very slider-heavy maps with SB at the end, so it's quite reasonable.

On the second screenshot there's one red less, and few more greens. Two remaining reds are stream maps with very unreasonable amount of missed sliderends (14 sliderend misses).