Improvement of hot scheduler when meets dimension conflict

[lhy1024]

Development Task

New Policy

• scheduler: enlarge the search space so that the hot-scheduler can still schedule under dimensional conflicts #4912
• scheduler: support the hot-region-scheduler to generate multiple operators at the same time #4931
• schedulers: skip uniform store to reduce schedule #5286
• Hot scheduler cannot handle some precise boundary conditions #5021
• statistic: collect both read and write pending influence at the same time #5521

Refactor

• scheduler: refactor calcProgressiveRank in hot-scheduler #4950
• scheduler: refactor solve in hot-region-scheduler #4952
• scheduler: redesigned progressiveRank of the hot-region-scheduler #4962
• Unify the length of Loads in hot statistics to 3 #5500
• *: move dim to statistics #5541

Fix Bug

• statistics: make aot more stable #4951
• scheduler: fix pending influence collect #5490
• statistics: hot peer cache contains unexpected data when cold peer and hot peer appear alternately #5478
• test: fix query statistic #5472
• scheduler: continuous move-hot-read-peer when exists evict-leader-scheduler #5492
• scheduler: fix write leader max zombie duration #5511
• scheduler: fix policy on any of in hot scheduler #5524
• scheduler: fix policy with write-leader in hot scheduler #5526

Misc

• statistics: show stores of peers in pd-ctl output #5330
• schedulers: Remove redundant judgment in hot-region #5369
• metrics: make hot split region more readable #5306
• metrics: add scheduling hot peer load #5525

Config

• scheduler: update hot region scheduler default config for v2 #5590
• scheduler: make SrcToleranceRatio and DstToleranceRatio enable for stddev #5546
• pd: add rank-formula-version config for hot scheduler pingcap/docs-cn#11566

[nolouch]

Please add more descriptions.

[lhy1024]

Performance improvement greater than 1% marked green.
performance regression greater than 1% marked red, stability evaluated by coefficient of variation, v2 greater than 0.05 and positive variation, marked red.
results are related to load and environment, marked yellow.
v2 is more balanced compared with v1, but the qps is not improved, marked blue; the scheduling convergence result is related to the split result, marked light blue.