multi-tenant support

[neverchanje]

We can consider to assign a "cache quota" to app when we support multi-tenant.
For example given with a specific SLA requirement, we can assign the memory size of block cache to each replica according to the following algorithm:

                                    lower qps require smaller cache
                                                    |
(data_size / partition_count / 5) *  (read_qps / partition_count / 100)
                |
   single partition data size, 1/5 is the memory/disk ratio

For a app that has 2000 read qps, 200 GB estimate data size, with 128 partitions,
the cache size is:

(200GB / 128 / 5) * (2000 / 128 / 100)  = 50MB

[neverchanje]

docs of hbase multi-tenancy: https://github.com/apache/hbase/blob/master/src/main/asciidoc/_chapters/ops_mgt.adoc#71-quotas