Implement maxOps limit for IndexSizeTrigger

IndexSizeTrigger can currently generate unlimited number of requested operations (such as split shard). Combined with the avalanche effect in SOLR-12730 this may cause a massive spike in the cluster load, and some of these operations may fail.

It's probably better to put an arbitrary limit on the maximum number of generated ops requested in one trigger event. This will delay some of the needed changes (thus leading to thresholds being exceeded to a larger degree) but it will increase the likelihood that all operations succeed.

A similar limit already exists in SearchRateTrigger.