Expose the size of the HT to Prometheus to allow tuning for less HT resize

Added metrics

kv_vb_ht_avg_size

kv_vb_ht_max_size

Background

Discussed with the team in the CBSE meeting yesterday. We have some customers which want to avoid the slowness experienced during HashTable resize at all cost, as their dataset grows and shrinks rapidly.

They could set their initial HashTable size to a large enough size to avoid resizing occurring, as the expense of higher memory usage.

To be able to choose the correct initial HashTable size, they can use the maximum size of any HashTable (vBucket), and we can expose that to Prometheus. Exposing the sizes of all HashTables will introduce too many metrics.