iterative pruner follow-up

[J-shang]

Need follow-up:

• check compression doc and update expired content.
• TaylorFOWeightFilterPruner, ActivationAPoZRankFilterPruner, ActivationMeanRankFilterPruner check these pruners if need multi-epoch training. (Workaround by batch, need refactor)
• SlimPruner check if we need support multi-iteration.
• AutoCompressPruner ADMMPruner SimulatedAnnealingPruner check the relationship between these pruner.
• test patch_optimizer_before() (callback)
• update ut
• statistics_batch_num disappear?
• LotteryTicketPruner need check optimizer

[J-shang]

• Refactor compress() in IterativePruner, unify support AGP, ADMM.
• Some pruners have different pruning conditions, not only sparsity, do we need to support?
• Slim, TaylorFOWeightFilterPruner, ActivationAPoZRankFilterPruner, ActivationMeanRankFilterPruner need real iterative. Related to above item.
• Add full test (l2filter, Taylor, NetAdapt, SimulatedAnnealing, AutoCompress, AMC, Sensitivity, ADMM, Lottery Ticket)

[J-shang]

• iterative -> (apply sparsity rate to layer & pruning) / (pruning by other conditions)