Attempting to delete tablets with massive channel history leads to oom

[vporyadke]

An attempt to delete a table on a development cluster led to the cluster being down - Hive is oom-killing any node it runs on.

The memory is in BS_QUEUE:

The source of the problem appears to be this:
https://github.com/ydb-platform/ydb/blob/main/ydb/core/tablet/tablet_req_delete.cpp#L44 - TabletReqDelete actor creates a request for every entry in tablet history, and the tablets have >1000 entries each.

1. We should limit the number of inflight requests in TabletReqDelete This should be fine, since it already does one message per group+channel pair. We should limit the number of inflight tablet deletions instead.
2. This might not be enough, we might need to also do some other changes in Hive,

   ydb/ydb/core/mind/hive/tx__delete_tablet_result.cpp

   Line 31 in aa7a166

   for (const TTabletChannelInfo::THistoryEntry& historyInfo : channelInfo.History) {

   - e. g. could this go over the commit size limit?