[CBSE] Reduce memory allocation for mutation queue

As seen in a customer case, with 10GB of memory quota for indexer, mutation queue got 1GB allocation. With slow storage drain rate and a full mutation queue, the heap usage was following:
1. 1GB for the queued up mutations
2. 700MB for the queued up TS(23K TS each having a size of 32k)
3. 700-800MB protobuf garbage.

The memory allocation for the queue can be reduced further to reduce the overheads.